AI tools and open source LLMs explained: what they do and how to use them, including free ones and ones that run on your computer

What it is

An open source platform for building AI-powered applications by dragging blocks onto a canvas: chatbots, multi-step workflows, and assistants that answer using the company's documents. It connects to hundreds of models, including GPT, Claude, Mistral, and Llama.

Who it is for

Teams that want an internal assistant that answers questions about company procedures using their own documents, or a workflow that starts automatically when an event happens, without opening a development project.

Where it runs

Online on Dify Cloud, with nothing to install. Or on your own server with Docker: you need at least 2 CPUs and 4 GB of RAM.

Difficulty

Easy in the online version. Installing it on a server requires a technician.

Cost

Online: free Sandbox plan to try it out (200 message credits, 5 apps, one user). Professional: $59 per month or $590 per year. Team: $1,590 per year. Enterprise: quote on request. Prices exclude VAT. On your own server: free Community edition, under a license based on Apache 2.0 that adds some conditions for commercial use.

Getting started

1. Create a free account at dify.ai and try the Sandbox plan.
2. Start from a ready-made model (chatbot or workflow) and connect the API key of the model you want to use.
3. If you need to keep it in-house, have a technician install the Community edition on a server using Docker.

What it is

n8n is a platform for building automated workflows and AI agents. You build them with a visual editor, and steps can include JavaScript or Python code. It connects different apps and moves and transforms data between them. The official site mentions more than 500 integrations, while the README on GitHub lists more than 1500.

Who it is for

Useful for people who repeat the same tasks across different apps every week, for example a small team that wants to automatically copy online form responses into a spreadsheet. Technical teams also use it to manage where the data for an AI assistant passes.

Where it runs

Online, through the cloud service managed by n8n (data on the cloud plans is stored on servers in Frankfurt, Germany). Or on your own computer or a server, by installing n8n with Docker: the Community edition is available on GitHub.

Difficulty

Medium: the cloud version is used from a browser with the visual editor, but installing it on your own computer or server requires Docker and some technical skills.

Cost

Cloud plans, with prices listed on the official site as billed annually: Starter €20 per month; Pro €50 per month; Business €667 per month, available only self-hosted; Enterprise on quote. A free trial is available with no credit card required, with limits (1,000 executions and 5 concurrent executions). The self-hosted Community edition is available on GitHub under a fair-code license (Sustainable Use License).

Getting started

1. Open n8n.io and click "Get Started" to start the free trial, no credit card required.
2. Choose a sample workflow from the workflow template library and try it with your test data.
3. To install it on your computer, follow the official Docker installation guide.
4. For questions, use the community forum linked in the documentation.

What it is

Claude Mods is an unofficial directory of mods for Claude Code, Anthropic's program used from the terminal. It currently lists 211 mods, each with its own listing and the commands to install it. Mods are plugins that run inside Claude Code and can show elements above the prompt, open panels, block or change tool operations, and keep secrets and sensitive data away from the model.

Who it is for

Useful for people who use Claude Code and want to see what is happening during their work, such as plan usage, subagent status, or file changes. Someone working on a long project might, for example, look for a mod that shows plan usage limits above the prompt.

Where it runs

On your computer. Mods work in the terminal and in the Code tab of the Claude Desktop app, with Claude Code 2.1.287 or later. The directory is browsed in a web browser at claude-mods.com.

Difficulty

Medium. Browsing the list is simple, but installing a mod means using the terminal and two commands, and each mod runs with the permissions of your Claude Code.

Cost

Price not stated on the official site for the directory. Each mod belongs to its author and is distributed under its own license. Some mods call a model using your plan, so they use up your usage.

Getting started

1. Check your version with claude --version and update Claude Code to 2.1.287 or later.
2. Open claude-mods.com, browse the list, and pick a mod by reading its listing.
3. Before installing, download the repository and check what the mod does with claude plugin validate, as explained on the installation page.
4. Run the two commands shown in the listing, then start claude and check the mod with the /plugin command.

What it is

Commonset is a platform for organizations that collects the AI skills, instructions, and workflows people create. For each item, it shows the owner, version history, and review status. An item can enter Commonset from the browser, through an integration, or via MCP, and the team can then find and reuse it.

Who it is for

Designed mainly for teams and companies that use several AI skills and want to keep colleagues from redoing the same work. A single person can also use it to organize their own skills, but the product is built for shared work. Example: the legal department uploads a skill for reviewing contracts, and the rest of the company can find it already approved, without rewriting it.

Where it runs

Online: used from the website and the web app (app.commonset.ai). For developers, an MCP interface and a REST API are also available. The official pages do not describe an installation on your computer or on your own server.

Difficulty

Medium: signing up is simple, but connecting Claude, OpenAI, Google, or GitHub and setting up permissions and approvals requires someone who knows the accounts and the organization.

Cost

The site shows a «Start free» button for signing up, but does not list plans or prices. Price not stated on the official site. For a team offer, you can request a demo through «Book a demo».

Getting started

1. Open commonset.ai and click «Start free» to create an account.
2. Bring a skill or instruction into Commonset from the browser, an integration, or MCP.
3. Assign an owner and check the security and integrity evidence before approving a version.
4. If you work in a team, check which platforms can connect (Claude, OpenAI, Google, GitHub) or request a demo.

What it is

The project is a collection of agents, commands, skills, rules, and hooks for Claude Code, the tool for programming with AI help. The files are published on GitHub, and there are dedicated agents for code review, security, and testing, for example. It is not a standalone program: you install it inside Claude Code, where it adds commands such as /tdd or /plan.

Who it is for

Useful for people who already use Claude Code to write software and want to start from a ready-made setup, for example to have their code checked by a dedicated review agent. People who don't code get little direct use from it, because the instructions use terminal commands and configuration files.

Where it runs

On your computer, with Windows, macOS, and Linux. The scripts are written in Node.js, so you need Node.js as well as Claude Code.

Difficulty

Medium: you need to install Claude Code, add the plugin or copy files into system folders, and replace the API key placeholders in the MCP server configuration files.

Cost

Free. The repository is released under the MIT license. The price of Claude Code, which you need to use it, is not stated on this page.

Getting started

1. Install Claude Code on your computer.
2. Open the Installation section of the GitHub page and choose either the plugin install or the manual install.
3. Check the install command before running it: the page shows the path affaan-m/everything-claude-code, which differs from the WorldFlowAI repository in this card.
4. Read the Shorthand guide, which the project lists as the starting point, before you enable hooks and MCP servers.

What it is

FindSkills is an online catalog of skills for AI agents, that is, modules that add abilities such as connecting to a database or a browser. The site indexes over 94,000 open source skills from GitHub, ClawHub, and the OpenClaw ecosystem, including MCP servers and skills for Claude Code. The site warns that skills are submitted by the community and not verified, so you should check them before letting them act.

Who it is for

Useful for people who use an AI assistant such as Claude or Cursor and want to add features to it without searching through many repositories. Example: someone who wants the assistant to read a web page or query a database can search the site for the right skill and install it.

Where it runs

Online, at findskills.org, to search and browse skills. On your computer, from the command line with npx findskills, or as an MCP server with npx findskills-mcp for Claude Desktop, Claude Code, and Cursor.

Difficulty

Medium: searching the site is simple, but connecting the agent through MCP requires using the terminal and editing or creating a configuration file.

Cost

Free. No sign-up is needed to search the site. Signing in with GitHub gives you a free key, free during the preview phase, that raises the daily search limits. The license of each skill depends on its source project.

Getting started

1. Open findskills.org and search for a skill using a keyword from the directory.
2. To search from the terminal, run npx findskills "web scraping" and follow the instructions.
3. To connect Claude Desktop, Claude Code, or Cursor, run npx findskills-mcp init and restart the program.
4. Optional: run npx findskills auth to sign in with GitHub and get the free key.

What it is

SkillsHub MCP is a site that collects 982 skills for Claude, the AI assistant from Anthropic. A skill is an instruction file (SKILL.md) that tells Claude how to carry out a certain type of task, such as creating a Word file or an Excel spreadsheet. The site sorts them by category, including documents, business, marketing, web development, and productivity, and lets you download each skill as a ZIP file. Plugins for Claude Code are also available.

Who it is for

Useful for people who already use Claude every day and want more precise results without writing the instructions from scratch. For example, someone who needs to merge several monthly PDF reports into a single file with bookmarks can find the dedicated PDF skill and use it with one natural-language request.

Where it runs

Online: you can browse the catalog and download skills from the site. On your computer: skills work in Claude Code and in the Cowork desktop app, installed in a folder on your computer (~/.claude/skills on macOS and Linux, C:\Users\NomeUtente\.claude\skills on Windows). In claude.ai, you can copy the text of the SKILL.md file at the start of the conversation, without automatic activation.

Difficulty

Medium. For skills, you only need to copy a folder into the right place and restart the app, but plugins require Claude Code and commands run in the terminal.

Cost

The site says all skills are free and downloadable without registration. There is also a paid area with extra content: the page lists €5 / for 30 days for 30-day access and, when paying in USDC, 5 USDC for 30 days (Basic plan) and 49 USDC for one year (PRO plan). The documentation page states that Claude Code requires an active Claude subscription.

Getting started

1. Open the skills catalog and choose the one you need, by category or by name.
2. Download the ZIP file and unzip it: inside you will find a folder containing the SKILL.md file.
3. Copy the folder into Claude's skills folder (~/.claude/skills on macOS and Linux), then restart Claude Code or Cowork.
4. Describe the task in natural language: Claude activates the skill on its own when it is needed.

What it is

It is an AI agent skill, meaning instructions that an AI assistant carries out with a single command. The command reads the project, writes the plan and the sharing text, and produces a video with music and animations, saved in the brag-output folder. The video is generated with the Hyperframes tool. The letsbrag.app site offers the same idea online: you paste your site's link and get a video of about 20 seconds with music, animations, and text ready to publish.

Who it is for

Useful for anyone who has just published an app, website, or software project and wants a short video to announce it. For example, someone launching the website of their small online shop who wants a video for Instagram or LinkedIn without knowing how to edit video.

Where it runs

Two ways. Online, at letsbrag.app, with nothing to install. On your computer, as a skill in Claude Code, Codex CLI, opencode, or other agents that support skills, with Node.js 22 or higher, FFmpeg in the PATH, and the Hyperframes CLI.

Difficulty

Medium. The online version needs no installation, while the computer version requires a terminal, Node.js, FFmpeg, and Hyperframes setup.

Cost

The code is free and open source under the MIT license. On letsbrag.app the video costs $19 once (launch price), with $29 listed as the later price.

Getting started

1. For the online version, open letsbrag.app, paste your site's link, and choose the format (16:9, 9:16, or 1:1).
2. For the computer version, in Claude Code run /plugin marketplace add latent-spaces/brag and then /plugin install brag@brag.
3. Install Node.js 22 or higher and FFmpeg, then check Hyperframes with the command npx hyperframes doctor.
4. In the project folder, ask the agent to run /brag, and you will find the video in brag-output.

What it is

Agent Reach is a free, open source program published on GitHub that installs as a command on your computer. It doesn't read sites on its own. It chooses, installs, and manages the right tool for each platform, and then the AI assistant uses it directly. It covers web pages, YouTube subtitles and videos, RSS feeds, GitHub repositories and, with some setup, Twitter/X, Reddit, Bilibili, and XiaoHongShu. The command agent-reach doctor shows which channels are ready and which are not.

Who it is for

It's useful for people who use an AI assistant such as Claude Code, Cursor, or OpenClaw and want it to read content the assistant can't see on its own. For example, you could ask the assistant to summarize a YouTube video from its link, or to search Reddit for how other users solved a problem.

Where it runs

On your computer, with Python 3.10 or later. The README doesn't list specific operating systems. It can also be installed on a server. In that case, the README says a proxy is needed, which isn't necessary on your own computer.

Difficulty

Medium: installation runs through the terminal and requires Python. For Reddit, Twitter, and XiaoHongShu, you need cookies or access to a browser you already have open, and the assistant can guide you through these steps.

Cost

Free. The README states that the tools are open source and the APIs are free. The only cost listed is a proxy of about $1 per month, which is needed only if you install it on a server. MIT License.

Getting started

1. If you use OpenClaw, first turn on the exec permission as described on the GitHub page. Then copy the install phrase with the link to the install.md file into your assistant.
2. Alternatively, install by hand from the terminal with pip install followed by the link to the project ZIP shown on the page, then run agent-reach install --env=auto.
3. Run agent-reach doctor to see which channels already work without any setup.
4. For Twitter, Reddit, or XiaoHongShu, ask the assistant to set up that channel and follow the steps it suggests.

What it is

It is an open source desktop layer from the W3BS project that lets AI agents you authorize use your real desktop. The agent can move the mouse, click, drag, type, use keyboard shortcuts, read what is on screen, take screenshots and see open windows. It connects to assistants such as Claude Code, OpenCode and Codex, and nothing goes through a cloud service. An icon in the taskbar or menu bar turns access on or off at any time.

Who it is for

It is useful if you already use an AI assistant and want it to handle repetitive tasks on your computer, such as opening a program and writing a note. Example from the site: asking Claude Code to open the text editor and write a short note. If you are not technical, read the permissions carefully before use, because an agent given the broad desktop level can act on the whole desktop.

Where it runs

On your computer, with supported systems. Linux with KDE Plasma and GNOME (x86_64 and arm64 processors), macOS 11 or later (Apple silicon and Intel), and Windows 10 or 11 (x64). Hyprland and W3BS DE are listed as installable and in testing. The site says no platform is certified yet.

Difficulty

Medium: installation runs from the terminal with a few commands, and you need an AI assistant already installed, such as Claude Code, OpenCode or Codex. You do not need to know how to program.

Cost

Price not listed on the official site. The site presents it as open source, with an Apache-2.0 license proposed for the original code, subject to dependency review.

Getting started

1. On macOS or Linux, open the terminal and run "curl -fsSL https://w3bs.org/install.sh | sh", then "w3bs install agentic-desktop-layer". On Windows, open PowerShell and run "irm https://w3bs.org/desktop/install.ps1 | iex".
2. Register your agent with "w3bs desktop agents add claude-code --preset broad". The "observe" preset only lets it look at the screen, while "restricted" limits it to the apps you name.
3. Connect the agent with "w3bs desktop setup claude-code --agent claude-code --http" and copy the command line that appears.
4. Turn access on from the taskbar icon or with "w3bs desktop control". To turn it off, use "w3bs desktop disable" or the emergency shortcut listed on the page.

What it is

It is a free skill for Claude Code, created by Jake Schincariol, that is activated with the /arena command. The tool launches up to 100 sub-agents that receive the same task, each with a different approach. The solutions compete in pairs, a judge scores them using a written rubric, and the losers leave the tournament. At the end it returns the surviving solution, the attacks it passed, and the number of rounds completed.

Who it is for

It is useful for people who use Claude Code to write code or text and are not satisfied with the first answer. For example, someone who needs to fix a failing test in a project file, or someone who needs to write the title of a pricing page and wants alternatives compared against each other.

Where it runs

On your computer, inside Claude Code, with Python 3.8 or later. You do not need to install anything else with pip.

Difficulty

Medium, because you need Claude Code installed and must copy a folder from the terminal, or install the skill as a plugin with two commands.

Cost

Free, under the MIT license. The official page says that Claude token costs remain the user's responsibility, and that the 100-agent version requires 595 sub-agent calls. There is also a --quick option, with 16 agents and 91 calls.

Getting started

1. Install Claude Code, because the skill does not work without it.
2. Copy the skills/arena folder into ~/.claude/skills/ as the README describes, or install the skill as a plugin with the commands on the page.
3. Check that the /arena command responds in Claude Code.
4. After an answer that does not convince you, type /arena or /arena --quick for a lighter contest.

What it is

claude-reflect is a plugin for Claude Code, the programming assistant you use from the terminal. It automatically records the corrections you make to Claude and the preferences you state, then, after you confirm, writes them into CLAUDE.md and AGENTS.md files. The /reflect-skills command searches past sessions for repeated tasks and suggests reusable commands.

Who it is for

Useful for people who use Claude Code often and get tired of repeating the same instructions. For example, if you have already corrected Claude several times about choosing a model, the tool can save that rule so Claude keeps it in mind in later sessions.

Where it runs

On your computer, as a Claude Code plugin. The official page says it supports macOS, Linux, and Windows, with Windows not requiring WSL. Requires Python 3.6 or higher.

Difficulty

Medium, because installation uses terminal commands and first requires installing Claude Code CLI, plus restarting the application.

Cost

Free. The project is released under the MIT license, and the GitHub page lists no prices for the plugin. Claude Code must be installed separately. Cost not stated on the official site.

Getting started

1. Install Claude Code CLI and check that you have Python 3.6 or higher.
2. In the terminal, run "claude plugin marketplace add bayramannakov/claude-reflect".
3. Run "claude plugin install claude-reflect@claude-reflect-marketplace" and restart Claude Code.
4. On first use, run /reflect. You will be asked to analyze past sessions to find corrections you already made.

What it is

It is open source software developed by IBM that acts as an intermediary between AI assistants and the services they use. It receives requests from protocols such as MCP, A2A, and REST or gRPC APIs, and routes them to the right tools, with access rules, service discovery, and monitoring. The official site describes it as a registry and proxy that connects tools, agents, and APIs in a single point for AI clients.

Who it is for

It is useful for people who manage several AI tools in a company or project and want to control them from one place. A concrete example is a team that connects an AI assistant to the office's internal data and wants to decide who can use which tools.

Where it runs

On your computer, installed through PyPI (with the uv tool) or through Docker or Podman. On a server, with Docker Compose, Kubernetes or OpenShift, Helm, and guides for cloud services such as Google Cloud Run, IBM Code Engine, AWS, and Azure.

Difficulty

A technical person is needed. You have to install and configure the software from the terminal, generate an access token, and connect it to your own AI tools.

Cost

Open source software under the Apache-2.0 license, so it is free to download and use under the terms of the license. The official pages checked do not list any paid plans. Price not listed as a paid plan on the official site.

Getting started

1. Read the official Quick Start guide on the project site, where the installation methods are listed.
2. Install the uv tool, listed as a prerequisite, then start the gateway with the PyPI command, or use Docker or Podman with the command from the guide.
3. Open the local address given in the guide (for direct installation it is http://localhost:4444) and sign in to the admin interface.
4. Generate an access token by following the guide, then register your MCP tools and try the first example.

What it is

It is a plugin for Claude Code that replaces the automatic summary used when a conversation gets too long. The system sends the list of tool calls and results to Jev, a TypeSafe service, and receives decisions on what to keep. Items no longer needed are removed or shortened, while the messages written by the user and the assistant stay intact and in order. The code is also available as an npm library, the JavaScript package manager.

Who it is for

Useful for people who use Claude Code for long sessions, with many files read and commands run, and who want to avoid losing details such as file paths or exact error messages in the summary. Example: a developer who has worked on a project for hours and uses the /compact command to free space without losing the name of the file they are editing.

Where it runs

On your computer, inside Claude Code. The plugin uses an early-access Claude Code feature (version 2.1.274 or later) that you turn on with a setting in a configuration file. The decisions are calculated by TypeSafe's online service, which requires an API key.

Difficulty

Medium: you need to turn on an early-access feature, add an API key to a settings file, and run a few terminal commands.

Cost

The code is free, under the MIT license. The Jev service from TypeSafe requires an API key. Price not verified: the official pricing page could not be checked.

Getting started

1. Get a TypeSafe API key and keep it private, without putting it in shared files.
2. In the file ~/.claude/settings.json, set CLAUDE_CODE_ENABLE_FUNCTION_HOOKS to 1 and add the key as TYPESAFE_API_KEY.
3. From the terminal, add the repository as a marketplace with: claude plugin marketplace add tamaratran/fast-jev-compaction
4. Install the plugin with: claude plugin install fast-jev-compaction@fast-jev-compaction, then restart Claude Code or run /reload-plugins.

What it is

Satori is a skill for Claude, meaning a set of instruction files that the assistant reads to change how it responds. The project blends eight traditions of thought, such as Stoicism and Buddhism, with some psychotherapy methods, to offer a deeper conversation. It is published on GitHub by MetcalfSolutions. It is not a therapist and does not make diagnoses. When a problem goes beyond conversation, it says so and points the person toward a professional.

Who it is for

It is useful for someone who wants to reflect on their habits over time, somewhere between I am fine and I need a professional. For example, someone who often starts projects and does not finish them can use it to understand what happens in those moments, without expecting a ready-made solution.

Where it runs

Online, in claude.ai, in the Skills section. On your computer, with Claude Code, by copying the skill files into the skills folder. No server needed.

Difficulty

Easy with the zip file, which needs no terminal and no code. A bit more computer familiarity is needed for the Claude Code route, which goes through the terminal.

Cost

Free. The GitHub page says Free, open-source and No subscription, under an Apache 2.0 license that requires attribution. The page does not say which Claude plan is needed to use skills, so the cost of Claude is separate.

Getting started

1. Download SatoriSkill-v5.zip from the Releases page on GitHub. No need to unzip it.
2. Open claude.ai, go to Customize, then Skills.
3. Click the +, choose Upload a skill, and upload the zip.
4. Open a new conversation and say what you want to talk about. The skill activates on its own.

What it is

AI-Cockpit is a Mac and iPhone app built by an independent developer in Switzerland. It shows the limits and spending of seven AI services, including Claude, ChatGPT with Codex, the Anthropic and OpenAI APIs, Kimi K3, OpenRouter, and xAI's Grok, with one card for each. On the Mac it lives in the menu bar and can also show active Claude Code sessions, while on iPhone there are Home Screen widgets.

Who it is for

It's useful for people who use several AI subscriptions or APIs every day and want to know how much credit is left before they get stuck. Example: someone with a personal Claude account and a work one can keep them on two separate cards, with different names.

Where it runs

On your computer: a Mac app that requires macOS 14 or later, in the menu bar. On your phone: an iPhone and iPad app that requires iOS 26, with Home Screen widgets; there is also a companion for Apple Watch. It is not an online service: data stays on the device, and the app does not require an account.

Difficulty

Medium, because OpenAI and Anthropic require an admin key, which you get from your account with the provider. The other services are simpler to connect.

Cost

CHF 4.00 / $3.49, one-time payment, according to the official page. There are no subscriptions or in-app purchases. The subscriptions and API credits the app reads remain your responsibility with the providers.

Getting started

1. Download the app from the Mac App Store or the App Store.
2. On first launch, follow the four-step assistant. Each step can be skipped.
3. To try it without connecting anything, open demo mode, which shows sample data.
4. Add keys only for the services you use, in Settings, under Account.

What it is

Context Mode is a service that sits between AI coding agents, such as Claude Code or Codex, and the model that runs them. It keeps raw data produced by the tools out of the assistant's working memory, so the model reads only the useful answer. It also remembers the work done in long sessions, so it does not disappear when the conversation gets shorter. The model is paid through your own Anthropic or OpenAI account.

Who it is for

It is useful for people who use an AI assistant every day to write code and see costs rise or lose track of long projects. Example: a developer who works for days on the same website and wants the agent to remember files, errors and decisions already made.

Where it runs

Online: the service runs on the Context Mode cloud. On your computer: the free Engine plugin works locally with the supported AI coding tools, and data stays on your machine unless you connect the Insight service.

Difficulty

A technical person is needed for the first setup, because installation uses terminal or agent chat commands. After setup, daily use requires no special commands.

Cost

Free plan: $0, 1,000 requests once, no card. Starter: $20 a month or $200 a year. Pro: $49 a month or $490 a year. Past the limit, Starter and Pro charge $2 per 1,000 requests (Starter) and $1 per 1,000 requests (Pro), up to a budget you set. Team: per seat, talk to us. Enterprise: yearly contract. The Engine plugin is free, under the Elastic License 2.0.

Getting started

1. Open the quick start guide at context-mode.com/docs/quick-start and create a free account with GitHub or Google.
2. Install the free Engine plugin in Claude Code with the commands /plugin marketplace add mksglu/context-mode and /plugin install context-mode@context-mode.
3. Restart Claude Code and run /context-mode:ctx-doctor to check that all tests pass.
4. Stay on the Free plan until the 1,000 requests run out, then choose Starter or Pro from the pricing page.

What it is

Fermat's Last Token is a plugin from Quotient Labs, a San Francisco company backed by Y Combinator. It runs alongside Claude Code, Anthropic's coding tool, and reduces the tokens used in a session by compressing context and tuning tool calls. You keep working as before. The site reports an average saving of 47% over 15 test sessions, with a quality score that went from 0.83 to 0.92.

Who it is for

It suits people who run long coding jobs with Claude Code and want to control the bill, for example someone refactoring a project with many files. Users who pay the API per use see the savings directly in dollars, while users on a flat-rate subscription get higher usage limits.

Where it runs

On your computer, from the terminal or through the Claude Code extension for VS Code. You need access to Claude through Anthropic or Vertex AI. The site lists AWS Bedrock as coming soon. Supported operating systems are not listed on the official site. The dashboard for your account and savings is online.

Difficulty

Easy if you already use Claude Code, because installation is a single command and your workflow stays the same. Some comfort with the terminal is still needed.

Cost

The site offers two plans: "$20/mo" for unlimited usage, or "10% metered", meaning you pay 10% of the savings you get. There is a 2-week free trial, listed as a limited-time offer through October 31. The site does not state a currency beyond the $ symbol or whether taxes are included.

Getting started

1. Open https://www.quotientlabs.com/ and check the free trial terms.
2. In the terminal, run: curl -fsSL https://downloads.quotientlabs.com/fermat/install.sh | bash && fermat login
3. Finish signing in through the browser that opens. The free trial starts there.
4. Start Claude Code as usual. To use it without Fermat, run claude --vanilla.

What it is

Langfuse is an open source platform for people who build apps based on language models. It records what happens with each request the app makes, lets you manage and update the instructions given to the model (the prompts), and helps evaluate the responses, including through manual reviews. It receives data from the app through SDKs for Python and JavaScript/TypeScript, or through an API, and lets you examine logs and user sessions.

Who it is for

It is useful for teams that develop a chatbot or AI assistant and want to understand why some answers are not good. Example: a company with a customer assistant can see which questions get wrong answers and fix the instructions given to the model.

Where it runs

Online, with Langfuse Cloud, managed by the Langfuse team. On your own computer, with Docker Compose, or on your own server (VM or Kubernetes). The self-hosted setup requires PostgreSQL, ClickHouse, Redis or Valkey databases, and S3 or equivalent storage. The documentation does not state minimum hardware requirements.

Difficulty

A technical person is needed to install it on your own server, because you have to manage Docker and several databases. The online version avoids installation and is the simplest choice for a non-technical user.

Cost

Online (Langfuse Cloud): free Hobby plan, no credit card required, with 50,000 units per month. Core $29 per month and Pro $199 per month, each with 100,000 units included and $8 per additional 100,000; Enterprise $2,499 per month. On your own server: open source under the MIT license, with no license costs.

Getting started

1. Go to cloud.langfuse.com and create an account on the free Hobby plan.
2. Try the interactive demo mentioned in the documentation first, without installing anything.
3. Connect your app with the Python or JavaScript/TypeScript SDK by following the documentation at langfuse.com/docs.
4. If you want to host it yourself, follow the Docker Compose guide at langfuse.com/self-hosting, preferably with a technical person.

What it is

It is a free, open-source command-line tool under the MIT license. It reads token counts from the files Claude Code, Codex and opencode already save on your computer, without reading your messages or the answers. It shows where the costs go, suggests fixes with their price in dollars, and applies them only if you accept.

Who it is for

Useful for people who use Claude Code, Codex or opencode often and want to know why their AI spending is high. Example: someone who picks up a long conversation after a few days and wants to know what it costs to have the assistant read it again.

Where it runs

On your computer, from the terminal, with Node.js (version 20.11 or later; 22.13 or later to read opencode history). Works with Claude Code, Codex and opencode 1.x. For your own API calls, one line of code is needed.

Difficulty

Medium: you use it from the terminal and must install Node.js. No programming is needed, but you have to type a few commands.

Cost

The command-line version is free, with no account, under the MIT license. Pro, the hosted version with a dashboard, is announced as coming soon; price not listed on the official site.

Getting started

1. Install Node.js on your computer, version 20.11 or later.
2. Open the terminal and type npm i -g optimaizr to install the tool.
3. Type optimaizr profile to see where your tokens go and which fixes are worth making.
4. Use optimaizr undo followed by the rule name to reverse a change you already applied.

What it is

roborev is a free program, released under the MIT license and developed by Kenn Software LLC. It runs in the background on your computer and, on every commit to a Git project, has an AI agent you have already installed review the code. It returns a result and, if it finds problems, a list of findings with severity and the file involved. You can read them in the terminal or in the browser, and send them back to the agent so it can fix them.

Who it is for

It is useful for people who write code with Claude Code, Codex, or another AI agent and want a second check before submitting changes. Example: someone who has an agent write the payment function for a small site and wants to see the findings before publishing it.

Where it runs

On your computer, with macOS, Linux, or Windows. You also need at least one coding agent installed on the same machine. No server needed.

Difficulty

Medium: installation is done from the terminal, and the project must be a Git repository. The site also suggests having your own agent complete the setup.

Cost

Free software under the MIT license. The official page does not state a price for the program. Reviews use the agents you already have installed, so the costs remain those of the subscriptions or keys you already pay for.

Getting started

1. Install roborev with one command: on macOS or Linux using the script or Homebrew, on Windows using PowerShell.
2. Go to your Git project folder and run roborev init, which installs the commit check and starts the program in the background.
3. Optional: run roborev agent-hook install to connect your coding agents.
4. Open roborev tui in the terminal or roborev ui in the browser to read the reviews.

What it is

It is an open source scanner from NVIDIA that analyzes skills for AI agents, such as those used with Claude Code, Codex, and MCP servers. It looks for risks like hidden instructions that try to override the assistant, data theft, and vulnerable dependencies. It accepts a folder, a SKILL.md file, a zip archive, or a Git repository, and returns a risk score from 0 to 100 with a report in terminal, JSON, Markdown, or SARIF format.

Who it is for

It is useful for people who install skills or plugins for AI assistants downloaded from the internet and want to check them first. For example, someone who finds a skill on GitHub for their assistant can run a scan to see whether it contains suspicious commands before using it.

Where it runs

On your computer, from the command line, with Python 3.12 or higher installed with uv or pip, or with Docker and no Python needed. The deeper analysis with an AI model also needs a key or account with the provider you choose (for example OpenAI, Anthropic, or NVIDIA build), or a compatible local model.

Difficulty

Medium: you use it from the terminal and need to install Python or Docker and set any API keys with commands.

Cost

Free: the software is released under the Apache 2.0 license. The official page does not list prices. If you use the AI model analysis, costs depend on the provider you choose, and prices are not listed on this page.

Getting started

1. Install uv or Python 3.12 or higher on your computer.
2. Run the command: uv tool install git+https://github.com/NVIDIA/skillspector.git
3. Try it on a skill folder with: skillspector scan ./my-skill/ --no-llm (static analysis only, no AI models).
4. For the AI model analysis, set SKILLSPECTOR_PROVIDER and the key for your chosen provider, then run the scan without --no-llm.

What it is

It is a public site that collects and sorts AI news, curated by Clairevue. It shows product launches, funding rounds, job openings, and events, and each item links to the original post, including posts from LinkedIn, X, and Hacker News. It can be read in English, German, and Spanish, and filtered by news type, source, country, and language.

Who it is for

Useful for anyone who wants to follow the AI sector without checking several sources every day. For example, someone looking for an AI job can filter job postings by country and open the company's post directly.

Where it runs

Online, in a browser, on the public feed page. It can also be connected to AI assistants through an MCP server by entering the address in the assistant's settings.

Difficulty

Easy: browsing the feed requires nothing, with no account or installation. Only connecting to assistants takes a few steps in the settings.

Cost

Browsing the public feed is free. The official page says API and MCP access is currently free, with limits: 1,000 requests per day (UTC), 30 per minute, and 10 new semantic searches per day. The official pages reviewed do not list paid plans. Update emails are announced but not yet available.

Getting started

1. Open https://signals.clairevue.ai/recent-signals and scroll through the feed: no account needed.
2. Use the filters for news type, source, country, or language to narrow the list.
3. To use it with an AI assistant, add an MCP connector with the address https://signals.clairevue.ai/mcp and choose OAuth authentication.
4. Sign in with your email and approve read access. Semantic search is optional.

What it is

Crash Test AI is a site that puts AI tools, assistants, and workflows through tests under conditions close to real use. A single person, Woody, runs it. He collects the evidence and describes what actually happens, including failures. Each published test gets a score out of 100 and a verdict (Pass, Mixed, or Fail), and it names the main point of failure.

Who it is for

Useful for people who need to choose an AI tool for work and want to know where it breaks. For example, someone deciding whether to hand source-finding for an article to a research assistant will find a test here with the problems that came up. Those looking only for quick reviews will instead find longer tests, sometimes based on months of use.

Where it runs

Online. The site is used from a browser, with nothing to install.

Difficulty

Easy, because each verdict highlights the score and the main point of failure. The pages are in English.

Cost

Reading the verdicts: the site does not state any cost. Newsletter: the site presents it as free. Private test of your own tool: £249 per tool, with a report and a verdict, according to the request page. The site says there is only one price, with no higher tiers. It is not stated whether the price includes taxes. Payment is due only after the slot is confirmed and the invoice is sent.

Getting started

1. Open the verdicts page and read the latest full test, for example the one on Dia Browser.
2. Read the page that explains how scores are assigned.
3. If you want to receive new verdicts, sign up for the newsletter on the site.
4. If you want to have a tool of yours tested, use the request page and state the tool's access and the use case.

What it is

MiroFish is an open source simulation engine that uses many AI agents, each with its own personality and memory. It accepts source material such as news articles, policy drafts, analysis reports, or novel texts, along with a natural language description of the question. It returns a forecast report and an interactive digital world, where you can talk with both the agents and the agent that writes the report.

Who it is for

It is useful for people who need to assess possible reactions and consequences, such as a communications office that wants to test a message before releasing it, or a writer who wants to imagine what comes next in a novel. The official pages do not report any tests of forecast accuracy, so use it to explore scenarios, not as a source of certainty.

Where it runs

On your own computer, with Node.js 18 or higher and Python 3.11 to 3.12, or with Docker. You also need access to a language model compatible with the OpenAI format and a Zep Cloud account for agent memory.

Difficulty

Needs a technical person, because installation requires using the terminal, filling in a file with API keys, and installing several tools (Node.js, Python, and uv).

Cost

The code is free and released under the AGPL-3.0 license. The official page does not list prices for the project itself. The README warns that using the language model consumes a lot and suggests starting with fewer than 40 rounds. Zep Cloud has a monthly free tier that the README considers sufficient for simple use. Language model costs and Zep plans are not listed on the MiroFish page and should be checked on the providers' websites.

Getting started

1. Open the official GitHub page and read the README file, available in English and Chinese.
2. Get an API key for a language model compatible with the OpenAI format, and create a Zep Cloud account.
3. Copy .env.example to .env, enter your keys, then install the dependencies with the command npm run setup:all.
4. Start with the command npm run dev and open http://localhost:3000 in your browser.

What it is

The service tracks odds across more than 30 sports and, before each event, publishes a pick with odds, stake, and reasoning. Results are closed automatically and stay visible, even when they are negative. The money is virtual, and ProphetAI does not take bets, because it is an information service and not a bookmaker.

Who it is for

Useful for anyone curious about how an AI handles simulated bets on real matches who wants to watch its record without spending anything. For example, a soccer fan can read the history of picks before deciding whether to subscribe. It is not suited to people looking for advice on playing with real money. When the site was consulted, the home page showed a virtual bankroll of £96.41 from a starting £100.

Where it runs

Online, in the browser. The diary can be read without logging in to an account. Paid tools and the members' Discord are reserved for subscribers, and there is also an API for those who want to connect their own scripts.

Difficulty

Easy: nothing needs to be installed and the site is in English, but the picks involve sports betting and should be read carefully with the risk in mind.

Cost

Free to read the log, the diary, and the public picks. Pro costs $9.99/month, with a standard price of $14.99/month, and includes the members' Discord and the analysis tools. API costs $19.99/month, with a standard price of $24.99/month. Both plans start with a 7-day free trial. The service is not open source.

Getting started

1. Open the public log and read the results, both the wins and the losses.
2. Read the daily diary to understand how the system reasons.
3. If the tools interest you, open the pricing page and choose Pro, with a 7-day free trial.
4. Before paying, read the cancellation terms and the 14-day right of withdrawal.

What it is

Crawl4AI is an open source program, created by unclecode and published on GitHub, that reads web pages and turns them into tidy Markdown text, suited to chatbots and AI agents. It accepts a page address and returns its content in Markdown, or structured data extracted with CSS or XPath rules, or with a language model. The cloud service is available with a free plan and paid plans, while the version you install on your own computer stays free.

Who it is for

Useful for people who work with web data: developers, researchers, or anyone preparing a knowledge base for an AI assistant. Example: copying the text of a competitor's pricing page and giving it to an AI assistant to compare prices.

Where it runs

On your own computer with Python, using the pip install command. On a server, using Docker. Online, using the cloud service with an API key. Minimum memory and operating system requirements are not stated on the official pages I reviewed.

Difficulty

Medium: installation goes through the terminal and uses Python, although a ready-to-use command-line tool (crwl) is included.

Cost

Program to install on your own computer or server: free, Apache-2.0 license. Infrastructure cost not stated. Cloud service, official prices in dollars: Free plan $0 with 1,000 credits per month, 10 requests per minute and 2 at the same time; Starter $10 per month with 10,000 credits; Builder $99 per month with 125,000 credits; Scale $499 per month with 750,000 credits. One credit is worth $0,001 and a simple page costs 0.5 credits. Taxes not stated on the pricing page. The page notes that prices may change during launch.

Getting started

1. Install the program with pip install -U crawl4ai, then run crawl4ai-setup to install the browser.
2. Try a first command: crwl https://news.ycombinator.com -o markdown. You will see the page in Markdown.
3. To use it without installing anything, create a free key on the crawl4ai.com site: the pricing page lists 1,000 free credits per month after email verification.
4. For advanced features, read the documentation at docs.crawl4ai.com.

What it is

An open source program that started at IBM Research. It reads PDF, Word, PowerPoint, Excel, web pages, emails and images, and returns ordered text: titles, reading order and tables stay in place. It also reads scanned PDFs.

Who it is for

People who work with documents full of tables, such as price lists, technical sheets or financial statements, and want to give them to ChatGPT or Claude without the numbers ending up in the wrong column. It also converts a whole folder of files at once.

Where it runs

On your own computer (Mac, Windows or Linux): documents do not leave the machine. It also connects to Claude Desktop as an MCP server, so Claude uses it when it receives a file.

Difficulty

Medium. You install it once from the Terminal, and it requires Python 3.10 or higher. After that, you use it with one command, or by asking Claude to convert the file.

Cost

Free, open source under the MIT license.

Getting started

1. Install it from the Terminal with pip install docling.
2. Convert a file with docling nome-file.pdf: the text version (.md) appears in the folder where you run the command.
3. To use it from Claude Desktop, add it as an MCP server by following the official guide.

What it is

FlagEmbedding is a Python library developed by the BGE project of the BAAI institute. It takes sentences or texts and returns their numerical vectors (embeddings), which make it possible to measure how similar two texts are. The project also includes reranker models for reordering results, examples, tools for training and evaluating models, and tutorials.

Who it is for

Useful for developers and researchers building an internal search engine or a system that answers questions using their own documents. Example: a company that wants to find, within its own manuals, the paragraphs closest to a customer's question.

Where it runs

On your computer, with Python installed via pip. It is not an online service. The models are published on Hugging Face and are downloaded by the code the first time they are used. System requirements not stated on the pages reviewed.

Difficulty

Needs a technical person, because installation and use happen through Python code and terminal commands.

Cost

Free. The repository is released under the MIT license. No paid plan listed on the official pages reviewed.

Getting started

1. Install Python and pip on your computer.
2. From the terminal, run: pip install -U FlagEmbedding (to use only the models) or pip install -U FlagEmbedding[finetune] (to also train them).
3. Copy the starter example from the GitHub page, which loads a BGE model from Hugging Face.
4. Try two or more sentences and calculate their similarity using the dot product of the vectors.

What it is

MinerU is an open source program developed by the OpenDataLab organization. It reads complex documents such as PDFs, scans, Word, PowerPoint and Excel files, and recognizes headings, tables and formulas. It returns the content in Markdown or JSON format, and in some cases also in HTML, LaTeX or DOCX.

Who it is for

It is useful for people who work with many documents and want to pull out their text in an orderly way, for example a researcher collecting data from scientific articles with tables and formulas, or someone building a knowledge base for an AI assistant from PDF manuals.

Where it runs

Online at the official site mineru.net (online use requires a login, according to the English-language GitHub page). On your computer, with Python from 3.10 to 3.14, by installing the mineru package. On a server, through Docker or with a local API service.

Difficulty

Medium, because the computer version requires installing Python and using the command line. The online site is easier to try.

Cost

Free to download. The repository uses the MinerU Open Source License, based on Apache 2.0 with additional conditions. For the online service at mineru.net, the price is not stated on the official site.

Getting started

1. Open the official page at mineru.net and try the online service if you want to test it without installing anything.
2. To install it on your computer, install uv (pip install uv), create an environment with Python 3.12 (uv venv .mineru --python 3.12), and run uv pip install -U "mineru>=4.0,<5".
3. Convert a file with the command mineru-kit parse documento.pdf -o documento.md --tier standard.
4. To use it from a browser window, start mineru-kit webui and open the local address listed in the documentation.

What it is

Onyx is an open source AI chat platform built for businesses. It connects to the apps a team already uses, such as documents and work tools, and searches their contents to answer questions. It can also search the web, run code in an isolated environment, and create documents to download.

Who it is for

It is useful for teams and companies that want an assistant that knows internal documents without anyone having to search for them by hand. For example, a support office asks the bot in Slack for an answer taken from the company's internal guide.

Where it runs

Online, with Onyx Cloud, no installation needed. Or on your own server, with Docker or Kubernetes. The Lite version needs less than 1 GB of memory and does not index documents. The site also mentions a desktop app.

Difficulty

A technician is needed to install it on a server, since it requires Docker or Kubernetes. The online version is easier to try on your own.

Cost

Onyx Community Edition is free under the MIT license, as stated on GitHub. On the official pricing page, the Business plan costs $20 per user per month, billed annually. The Enterprise plan is priced on request, to be discussed with sales. The site also offers a free trial (Start Free Trial).

Getting started

1. Go to cloud.onyx.app and sign up for the free trial to try the interface without installing anything.
2. If you want to install it on your own server, read the installation guide at docs.onyx.app. The project lists the command curl -fsSL https://onyx.app/install_onyx.sh | bash.
3. Connect the apps your team uses, choosing from the available connectors.
4. Check with your team which documents and permissions to enable before sharing access.

What it is

pgvector is a free extension for PostgreSQL, a widely used database. It lets you store vectors, meaning lists of numbers that describe text, images or other content, alongside regular data. With an SQL query, it returns the rows closest to a given vector, based on several distance measures. The project is open source and its code is on GitHub.

Who it is for

It is useful for people who build an app or website with a PostgreSQL database and want search by meaning to work, for example finding the documents most similar to a question or the products similar to the one the user is viewing.

Where it runs

On computers with supported systems: Linux, Mac and Windows, with PostgreSQL 13 or later. It can also be installed with Docker, Homebrew, APT, Yum, conda-forge and other package managers. On a server, it works with PostgreSQL databases managed by many hosting providers; the GitHub page notes that some of these already include it.

Difficulty

A technician is needed. Installation requires compiling the code or using a package manager, and use happens through SQL commands on an existing PostgreSQL database.

Cost

Free. The software is distributed under an open source license that, as the project's LICENSE file says, allows use without paying any fee. The GitHub page does not list paid plans or prices.

Getting started

1. Check that you have PostgreSQL 13 or later, or install it from the official site or your system's package manager.
2. Install pgvector: on Linux or Mac, download the code from GitHub, then run make and make install; on Windows, follow the instructions for nmake; or use Docker.
3. In each database where you want to use it, run the CREATE EXTENSION vector command.
4. Create a table with a column of type vector and insert vectors. Example instructions are in the README on the GitHub page.

What it is

Cover Your Tracks is an online test from the Electronic Frontier Foundation (EFF). It shows how trackers see your browser, including details such as screen resolution, language, time zone and installed fonts. At the end, it gives a report on how well your browser blocks trackers and how unique it looks to websites. According to the site, it collects only anonymous data.

Who it is for

Useful for anyone who wants to know how easily their browser can be recognized online, such as someone who uses an ad blocker and wants to check if it works. You can run the test in one browser and then in another, and compare the two reports.

Where it runs

Online: you use it from your browser at coveryourtracks.eff.org, with nothing to install.

Difficulty

Easy, because you only press the "Test Your Browser" button on the home page. No technical settings are needed.

Cost

The official site lists no price for the test. Its pages include a link to donate to EFF, but no cost for using the tool is listed.

Getting started

1. Open coveryourtracks.eff.org in the browser you want to check.
2. Press the "Test Your Browser" button and wait for the test to finish.
3. Read the summary on tracker protection and browser uniqueness.
4. For details, scroll the results page and open the detailed view.

What it is

It is an open source Docker image from the offen project that creates backups of Docker volumes, meaning the data that containers use. It runs alongside containers that are already running and can save the copies to a local folder or to S3, WebDAV, Azure Blob Storage, Dropbox, Google Drive, or an SSH server, including several destinations at once. It can run backups on a schedule or one at a time, encrypt the copies with GPG, delete the oldest ones if configured to do so, and send notifications about the outcome of each run.

Who it is for

Useful for people who use Docker to run applications and want copies of their data in a safe place. For example, someone who keeps a database in a container can set up a nightly copy to a folder on the server or to Google Drive.

Where it runs

On the computer or server, with the Docker CE engine, which is the one the project says it tests on. Other compatible engines may work, but without support guarantees. The documentation also lists arm64 and arm/v7 architectures and use in Docker Swarm mode.

Difficulty

Medium, because you need to know how to write a Docker Compose file and enter the settings and credentials for your chosen storage service in a configuration file.

Cost

Open source software under the MPL-2.0 license. The GitHub page and the official documentation do not list prices. The project also suggests a one-hour consultation, which you can book by becoming a sponsor.

Getting started

1. Read the official documentation at offen.github.io/docker-volume-backup, beginning with the quick start guide.
2. Add the backup service to your Docker Compose file and mount the volumes you want to save. The project recommends pinning a specific image version instead of using latest.
3. Create the configuration file with the backup destination and the credentials for your chosen storage service.
4. Start the containers and check the logs of the first run to confirm that the backup was created.

What it is

Legible is an online software platform for compliance of products with AI features. It keeps a list of AI systems, models, and vendors, rates the risks of each feature, tracks regulatory obligations, and collects evidence of controls. It then prepares documents to show auditors and customers. The example on the site includes the European AI Act, GDPR, HIPAA, ISO 27001, and SOC 2. It is in early access: to try it, you join a waitlist or request a demo.

Who it is for

It is useful for software companies that sell products with AI features in regulated industries, such as healthcare, and need to show customers and auditors how they manage risks and controls. Example: a company that sells an assistant to clinics for summarizing patient visits and must document what it does with patient data.

Where it runs

Online, as a web service. For now, access is through a waitlist or a demo.

Difficulty

Medium: you need people who know the product's compliance obligations. Access requires a 20-minute presentation with a specialist.

Cost

Cost not stated on the official site. No credit card is needed for early access.

Getting started

1. Open legible.global and choose Get early access or See Legible in action.
2. Fill out the form, indicating your role, to join the waitlist.
3. If you request the demo, it includes a 20-minute presentation with a compliance specialist.

What it is

rizzo-pii is an open source program created by Simone Rizzo and the Rizzo AI Academy. It finds personal data in a text, such as names, codici fiscali, partite IVA and cadastral data, and replaces it with placeholders like [CF_1]. The text with placeholders can be sent to ChatGPT, Claude or Gemini, and the reply is brought back locally with the original data.

Who it is for

Useful for law firms, accountants, notaries, and anyone who must comply with the GDPR and wants to use an AI on confidential documents. Example: a lawyer who wants a contract summarized without the clients' data ever reaching the cloud service.

Where it runs

On your computer. There are installers for Windows (.exe files), for Mac with Apple Silicon chips (.dmg files), and for Linux x86_64 (AppImage). The model runs on the CPU, without a graphics card and with about 0.5 GB of RAM.

Difficulty

Easy, because the desktop app is ready to use and requires no Python or configuration. Those who prefer can have an AI agent install the program using the text provided on the page.

Cost

Open source under the MIT License. Cost not stated on the official site.

Getting started

1. Open the download page and choose the file for your system (Windows, Mac with Apple Silicon, or Linux).
2. Install the program: on Windows, run the .exe file; on Mac, drag the app into the Applications folder; on Linux, make the AppImage executable with chmod +x and launch it.
3. Anonymize the text inside the app: the placeholder dictionary stays on your computer.
4. Copy the anonymized text into your AI assistant, then paste the reply back into the app to restore the real data.

What it is

Floot is a platform that builds apps and websites from what you describe in chat. It works inside Claude, ChatGPT, or Cursor, which you use as an assistant, and adds the database, the server-side technical part, user logins, and hosting on its own. The result is an online app with its own address, which can also be published as a phone app. The site says Git, a terminal, and build credits are not needed.

Who it is for

Useful for people who have an idea for an app or website but don't know how to code. For example, a small bakery can set up a system to take orders online, with a product list and a customer form, without hiring a developer.

Where it runs

Online. Use Floot at floot.com and connect it to Claude, ChatGPT, or Cursor as a connector. It is not installed on your computer.

Difficulty

Easy: you work by describing in plain language what you want. Only publishing to the app stores for mobile apps requires Apple and Google developer accounts, and the site warns that Floot builds the app but you must have the accounts yourself.

Cost

Free $0/mo: 100 build actions per day, 2 mobile apps per month, no AI credits, and no hosting included. Pro $25/mo: 1,000 actions per day, 5,000 AI credits per month, $10/mo of hosting, 12 mobile apps per month. Power $100/mo: 5,000 actions per day, 15,000 AI credits per month, $25/mo of hosting, unlimited mobile apps. With annual payment, the site lists 2 months free. The site does not list VAT. The site says Google Play requires a paid plan. Floot is not open source.

Getting started

1. Open floot.com/en and log in or sign up from the login page.
2. Connect Floot to Claude or ChatGPT using the official links on the main page, or add it to Cursor.
3. Describe the app you need in chat, then ask for changes one at a time until it is the way you want it.
4. Publish the app. For mobile apps, first check that you have Apple and Google developer accounts.

What it is

It is a free package of eleven skills for Claude, Anthropic's assistant, created by Jake Schincariol. Each skill handles one step: studying an existing app, planning its structure, redesigning it, building it, testing it, comparing it with the original, reading user complaints, changing its name and look, preparing the sales page, and publishing it. You start from an app's address, and the results go into a replica/ folder in your project, with a map of the screens, a list of features, and a work plan. The tool recreates the app's functions and way of use, not its code, logos, or content.

Who it is for

It is useful for people who have an idea for an app similar to an existing service, such as a booking system where you share a link and people pick a time slot. The first step, the study skill, shows how many screens and features are needed before you decide whether to start.

Where it runs

Online, inside Claude in the browser, by pasting the repository link. On your computer, with Claude Code, or by copying the folders into your own Claude workspace. The support tools are Python scripts, so you need Python 3.8 or a newer version.

Difficulty

Medium: basic use means pasting a link into Claude, but installing it as a plugin or copying the folders requires Claude Code, the command line, and Python.

Cost

Free. The repository is released under the MIT License, so you can use, modify, and distribute it. The page does not ask for registration or an API key. You do need access to Claude, however, which is a separate service not described on this page, so Claude's cost is not stated here.

Getting started

1. Open the GitHub page for the repository and read the Install section.
2. The simplest method is to paste the repository link into a Claude chat, followed by the phrase “install skill,” as the page indicates.
3. If you use Claude Code, you can add the repository as a plugin with the two commands listed on the page.
4. Start with the first tool, replica-recon, using the address of a public app that interests you. Use only public pages and your own account.

What it is

It is a collection of DESIGN.md files, each dedicated to the design of a well-known site such as Stripe, Notion, or Apple, published by the VoltAgent team on GitHub. Each file is plain text that describes the colors, fonts, spacing, components, and visual rules of that site, taken from public websites. Nothing to install and no data to send: you copy the file into your project and the assistant reads it to design the pages.

Who it is for

It helps people who build sites or apps with an AI assistant and want a consistent look without knowing design. For example, someone who wants a sales page in the style of a payments brand can copy that brand's file into the project folder and ask the assistant to build the page in that style.

Where it runs

Online, on the project's GitHub page and on the getdesign.md site, where you can browse and download the files. You then use the file on your own computer, inside a project folder that works with an AI coding assistant.

Difficulty

Medium, because you need a project folder and an AI coding assistant such as Claude Code or Cursor, tools that require some basic familiarity with computers.

Cost

The GitHub repository is open source under the MIT license, so it is free. The getdesign.md site also offers a Catalog Pass at $99 per month (on the site, "$99/mo"), with monthly renewal and cancellation at any time. The Pass includes 45 selected files, not the full catalog. Price for requesting a private DESIGN.md not stated on the pages reviewed.

Getting started

1. Open the project's GitHub page and scroll through the list of brands grouped by category.
2. Open the DESIGN.md file for the brand you like and copy the text or download it.
3. Save it as DESIGN.md in the main folder of your web project.
4. Ask your AI assistant to build a page that resembles that design, citing the file.

What it is

Impeccable is an open-source tool created by Paul Bakaus that gives your AI coding assistant rules and commands for web interface design. It offers 24 commands, such as polish for refinement, critique for review, and audit for technical checks, plus 61 automatic checks that look for the most common flaws in AI-generated designs, such as default fonts or cards nested inside other cards. You tell it what you want to improve on a page or component, and it either edits the code or reports what's wrong.

Who it is for

It's useful for people who build websites, landing pages, or dashboards with an AI assistant and want a more polished result. Example: you had an AI generate a pricing page, you find it too cluttered, and you run the command to make it easier to read and tidier.

Where it runs

On your computer, inside the project folder, alongside an AI coding assistant. The official site cites Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, and others. Installing with the npx command requires Node. There is also a Chrome extension that flags design issues within the open page.

Difficulty

Medium: you need an AI coding assistant and some basic comfort with the terminal for the install command. After installation, you type the commands into the chat as plain text.

Cost

Open source under the Apache 2.0 license, as stated on the official GitHub page. The official site shows no pricing page or paid plans. Cost not stated on the official site.

Getting started

1. Open the terminal in the project folder and run npx impeccable install, then follow the on-screen instructions.
2. Restart your AI coding tool. If it asks you to approve hooks or trust the folder, complete that step.
3. In the chat, type /impeccable init to explain who the project is for and what it's for.
4. Try a command on a page, for example /impeccable critique followed by the page name.

What it is

Penpot is an open source design platform built for teams working on digital products. It lets you create interfaces, interactive prototypes, and design systems with components and shared libraries, and it uses CSS Grid and Flex layouts. The site says it supports open standards such as CSS, HTML, SVG and JSON, and that the code can be inspected and integrated with the design.

Who it is for

Useful for interface designers, developers and product teams that want to work on the same file. Example: a designer draws an app screen, a developer copies its CSS code, and a colleague comments on the version in the same project.

Where it runs

Online, in the browser, through an account at design.penpot.app. Or self-hosted, meaning installed on your own server. The technical requirements for self-hosting were not verified on this page, so check them on the official self-host page.

Difficulty

Easy to use online, since all you need is an account and a browser. A technician is only needed if you choose to install it on your own server.

Cost

Professional plan: $0 user/month, with full features and no limits stated in the plan description. Unlimited plan: $7 editor/month, with a cap of $175 per month regardless of team size. Enterprise plan: $25 member/month. Private Server plan: 50k $ year. Prices in US dollars. Discounts are available for nonprofits, schools and open source projects. The core code is released under the MPL-2.0 license.

Getting started

1. Go to penpot.app and choose "Sign up, it's free" to create a free account.
2. Open a new project and create an empty file from the Penpot panel.
3. Browse the Libraries & templates section of the site to find libraries and templates you can start from.
4. If you want to install it on your own server, read the official self-host page before you begin.

What it is

Realtime Colors is a free tool that applies a color palette and two fonts to a sample website page, so you can see right away how it looks. You start with two neutral colors for text and background, then choose the primary, secondary, and accent colors. You can search for a font among Google Fonts or the ones installed on your computer, and at the end you export the result as CSS, SCSS, Tailwind, and other formats.

Who it is for

Useful for UI/UX designers, web developers, and anyone who manages a brand's visual style. For example, someone redesigning a client's website can show them the page with the chosen colors before they approve them.

Where it runs

Online, right in the browser, at realtimecolors.com. It is also available as a plugin for Figma and as a theme editor for VS Code.

Difficulty

Easy, because you only need to pick colors with the fields and buttons in the toolbar, and nothing has to be installed.

Cost

Free. The site says "100% free!" The pricing page lists the Basic plan as "Free" and the Pro and Enterprise plans as "$0.00 / month." The site's code is under a CC BY-NC-ND license. According to the FAQ, the generated colors can also be used for commercial purposes.

Getting started

1. Open realtimecolors.com.
2. Choose the text and background colors, then the primary, secondary, and accent colors.
3. Check the contrast indicator next to each color: red, yellow, or green.
4. Type the name of a Google Fonts font in the Headings and Body fields, then click Export to download the result.

What it is

Taste Skill is a set of instructions for AI coding agents, such as Claude Code, Codex, and Cursor, developed as an open source project by Leon Lin and blueemi. It gives the agent an artistic direction, real images, and control of the page in a browser, so it does not produce the usual standard page. It receives a project description and returns websites or app screens, built with the design language that the tool indicates.

Who it is for

It is useful for people who want an AI agent to create a landing page or app with a personal look. For example, someone who opens a small bar and asks their agent for an introduction page, without getting a design that looks like a thousand others.

Where it runs

On your computer. The tool runs with Node 22 or later and with Chrome, Edge, or Chromium installed. It works with Claude Code, Codex, Cursor, Gemini CLI, and other agents that read SKILL.md files. Finding images requires an internet connection.

Difficulty

Medium, because installation is done from the terminal and requires Node 22 or later and a compatible browser.

Cost

Free and open source under the MIT license. The official pages do not list paid plans for the tool. Support through GitHub Sponsors is optional. The tool does not require API keys.

Getting started

1. Install Node 22 or later and Chrome, Edge, or Chromium.
2. In the terminal, run: npx skills add https://github.com/Leonxlnx/taste-skill --skill "design-taste-frontend"
3. If you use Claude Code, you can instead type /plugin marketplace add Leonxlnx/taste-skill and then /plugin install taste-skill@taste-skill.
4. Ask your agent for a page or a screen, describing the project, the audience, and the mood you want.

What it is

DitherUI is a collection of animated React components, blocks (full screens), and interfaces for AI agents, such as chat and reasoning steps. You install it with the shadcn/ui command-line tool or by copying the code by hand. The source code ends up in your project files, so you can edit it freely. The site uses React 19, TypeScript, Tailwind CSS v4, Motion, and shadcn/ui.

Who it is for

Useful for people who build websites or apps with React and want well-made interfaces without designing every screen from scratch. Example: a developer who needs to build a page for booking appointments can start from the Booking Scheduler block on the site.

Where it runs

On the developer's computer, inside a React project (Next.js, Vite, TanStack Start, or similar) with Tailwind CSS and shadcn/ui. The component catalog is browsed online, with live previews.

Difficulty

A technician is needed: you must have a configured React project and use the command line to install the components.

Cost

Free plan: $0 forever, with the source code for the components in the documentation and use in personal projects. All-Access plan: $50 one-time payment, with all components, blocks, and templates, including future ones, and use in unlimited projects. The site does not state an open source license.

Getting started

1. Create or open a React project with Tailwind CSS and shadcn/ui
2. Add the @ditherui namespace to the components.json file, as shown in the installation guide
3. For All-Access components, create an API key on the profile page and set it as DITHERUI_API_KEY
4. Choose a component in the documentation, copy its command, and run it in the terminal; free components can also be installed from their address, without signing up

What it is

Glide Data Grid is a library for React programmers, developed by Glide, that builds data tables inside a website or web app. The cells are drawn on a canvas, so the table can handle millions of rows without weighing down the page. The programmer specifies the columns, the number of rows, and a function that returns the content of each cell. The resulting table offers cell editing, resizable columns, and selection of rows and cells.

Who it is for

It is useful for people who build a web app with React and need to show long lists, such as a management system with thousands of orders or an online spreadsheet. It is not meant for people looking for a ready-made program to use.

Where it runs

In the browser, inside a React app developed on your computer and installed with npm.

Difficulty

Requires a technical person, because you need to know how to write React code, install packages with npm, and connect your own data to the table.

Cost

Free and open source under the MIT license, also usable in commercial projects.

Getting started

1. Check that you have a React project, version 16 or higher.
2. Install the package with the command npm i @glideapps/glide-data-grid and the dependencies lodash, marked, and react-responsive-carousel.
3. Import the library's CSS file and add the DataEditor component to the page.
4. Define the columns and the function that supplies the data for each cell; the official quick start guide shows a complete example.

What it is

Pele is a small open source library that turns text written in the Mermaid language into SVG images. Mermaid is a way to write flowcharts in a few lines of text. Pele takes that text and returns an SVG, meaning a vector drawing that can be placed on a web page. The site says it has no external dependencies and is about six times smaller than Mermaid.js. Colors are changed with CSS variables.

Who it is for

Useful for developers who build websites or apps and want quick diagrams that match the site's visual theme. For example, someone writing an online manual who wants the diagrams to change color with the page's light or dark theme. The site says it was created for the Obsidian note-taking app.

Where it runs

Online: the site pele.run has a playground where you can try the code without installing anything. On your computer: you install it in a JavaScript project with npm, or load it from a CDN in a web page. It is not a standalone program to open with a double click.

Difficulty

A technical person is needed: to use it in your own projects, you must know how to write JavaScript and use npm. Trying it in the playground, however, is within reach of anyone who can copy and paste text.

Cost

Free. The site states that Pele is open source under the MIT license. On the home page, I did not find any paid plans or prices.

Getting started

1. Open the playground at pele.run/playground and edit the example diagram to see the result.
2. In your JavaScript project, install the library with the command npm install pele.
3. Pass the Mermaid text to the render() function and insert the returned SVG into the page.
4. To change colors and fonts, set the CSS variables listed on the Theming page at pele.run.

What it is

React Bits is a free collection of animated components for React, the technology used to build websites and web apps. The project was created and is maintained by David Haz, and its code is public on GitHub. Each component has a live demo with controls to change its values, and it installs as source code that stays in your project, in JavaScript or TypeScript, with CSS or Tailwind.

Who it is for

It is useful for people who build personal sites, portfolios or product pages with React and want animated text, moving backgrounds or buttons with effects without writing everything from scratch. Example: someone who wants the homepage title to appear letter by letter.

Where it runs

Online: in the browser you can try the demos and five free tools for creating backgrounds, mockups, logos, image effects and sounds. On your computer: components are added to a React project through the terminal, using the shadcn or jsrepo command, or they are copied by hand into the project. The pages I read do not list specific system requirements, but you need a React project.

Difficulty

Medium: you need a React project already running and some familiarity with the terminal, because installation happens with a command.

Cost

Free for the components and the tools. The site says the library will stay free forever, and the GitHub repository lists the MIT license with Commons Clause, which is free for personal and commercial use. There is also React Bits Pro, a paid option with lifetime or annual plans: the prices are not written on the pages I read, so I am not reporting them. Check pro.reactbits.dev for current figures.

Getting started

1. Open reactbits.dev, choose a component and try its demo.
2. If you have a React project, open the terminal in its folder.
3. Run the command shown on the installation page, changing the component name and the variant (JS or TS, CSS or TW).
4. Import the component into your file and display it on the page.

What it is

Caddy is an open source web server written in Go, developed as a project connected to ZeroSSL. It receives visitor requests and responds with files, pages, or by forwarding traffic to other programs. It turns on HTTPS by itself, gets and renews security certificates, and reads its settings from a simple file called Caddyfile.

Who it is for

It is useful for people who run a website or a small online project and want HTTPS without studying certificates. For example, someone with a blog on their own domain who wants it reachable over https without setting up anything complicated.

Where it runs

On your own computer or on a server, with Linux, Windows, macOS, FreeBSD, and other supported systems. It is not an online service: you install it as a program. It is also available as a Docker image and as a package for some Linux distributions.

Difficulty

Medium: installation is simple, but putting it online requires a domain, ports 80 and 443 reachable from the internet, and a basic familiarity with the terminal.

Cost

Free. The project is open source under the Apache-2.0 license. The official pages checked do not list any paid plans for the program.

Getting started

1. Open the releases page on GitHub and download the file for your system, or install with the package manager you use (for example, Homebrew on Mac or Docker).
2. Put the program in a folder included in your system's command path, so you can run it from anywhere.
3. Read the "Caddy basics" guide in the official documentation.
4. Write your first Caddyfile with the site name and start Caddy by following the guide.

What it is

DayClerk is a website that prepares a complete marketing campaign from a short description. Before writing the copy, it maps out how different audience groups might react, where they might drop off, and which words would persuade them. It returns landing pages published to a live web address, downloadable HTML email copy, social posts, and ads for Meta, Google, LinkedIn, and TikTok. It does not schedule posts, send emails, or run A/B tests, so publishing is up to you.

Who it is for

Useful for small business owners who run their own campaigns, agencies that manage several clients, and creators who promote other people's products. Example: the owner of a clothing store who wants one page for regular customers and another for people who don't know the store yet.

Where it runs

Online: you use it in your browser at dayclerk.com, with nothing to install. Pages are published to a live web address, while emails are downloaded as HTML to send through your own email service.

Difficulty

Easy: you work from a brief written in plain language and an editor where you can change headlines, text, and images without touching code. You only need to know what your offer is.

Cost

Explorer plan: $0 to start, no credit card required. Creating a page is free, but publishing it costs $29.98 once per campaign, with no subscription. Paid plans: Growth Engine $29/mo, Pro $59/mo, Agency $149/mo. The site says paying annually saves 20% on all paid plans. The site also offers a free demo, with no account needed.

Getting started

1. Open dayclerk.com and try the free demo, which doesn't require an account.
2. Write your brief: what you offer and the groups of people you're targeting. The free plan allows 2 groups.
3. Review the generated pages and copy, and edit them in the editor.
4. To publish a page, pay $29.98 once per campaign.

What it is

DayClerk is an online service that reads a business's website and points out what could be costing it leads. Then it writes and builds a sample marketing page designed for that person's audience. The demo asks for the website address and a few answers to the Clerxy assistant. The result is an analysis and a test page, not a published campaign.

Who it is for

Useful for small shop owners, service providers, and startups that don't have a marketing department. For example, a local business owner who wants to understand why the contact form on their website gets so few inquiries.

Where it runs

Online, right in the browser, with no account and nothing to install.

Difficulty

Easy, because the demo runs in the browser and requires no installation or technical skills.

Cost

Free demo, no account, no credit card, and one free run noted on the official site. Price to publish the page: cost not stated on the official site.

Getting started

1. Open the dayclerk.com/demo page in your browser.
2. Type your website address in the box.
3. Answer the questions the Clerxy assistant asks you.
4. Read the analysis of the problem areas and the sample page built for your audience.

What it is

It is a free collection of eleven skills for Claude, published on GitHub by Jakeschincariol under the MIT license. Each skill covers one part of the channel work: it writes the script starting from 21 hook formulas and scores the hook, checks the title and thumbnail as a pair, pulls a list of cuts from the transcript, reads the audience retention file to see where viewers leave the video, and finds Shorts inside a long video. It accepts texts, transcripts and retention exports, and returns drafts and guidance. It publishes nothing: you upload the video yourself.

Who it is for

It is useful for people who run a YouTube channel on their own and want to prepare videos with fewer attempts. Example: someone who has the transcript of a long video and wants to know which silences to cut and which part to use as a Short.

Where it runs

On a computer with Claude Code, using the six Python tools that require Python 3 and no other dependencies. Online, by pasting the SKILL.md file into a chat with Claude: this way the skills work, but the Python tools do not.

Difficulty

Medium: installation involves copying folders from the terminal with Claude Code, or pasting a link inside Claude. The Python tools require a computer with Python 3 installed.

Cost

Free, under the MIT license. The GitHub page does not state the cost of Claude, which is needed to use the skills: prices not stated on this page.

Getting started

1. Open the project's GitHub page and read the README file.
2. In Claude, paste the repository link and write: install this skill, then check that /yt-script works. Or, with Claude Code, copy the skills/yt-* folders into ~/.claude/skills/.
3. Fill in the templates/voice.md file by copying it to ~/.claude/youtube/voice.md, or ask Claude to create it from three of your videos.
4. Try the first command, for example /yt-script with a video idea, and check the result before using it.

What it is

An English-language article from the blog of the marketing agency Creative For More, published 17/10/2026, with product information updated to September 2026. It explains four things these tools measure (impressions, citations, brand mentions, and AI traffic) and separates data supplied directly by the platforms from data gathered with test prompts. It compares eight tools, including Google Search Console, Bing Webmaster Tools, Ahrefs Brand Radar, Semrush, Profound, Peec AI, Otterly.AI, and GA4, along with their limitations.

Who it is for

Useful for marketing professionals and website managers who want to decide which tool to choose for visibility in AI search. For example, a small team working in a single market can start with the free reports from Search Console and Bing Webmaster Tools, add GA4, and run a manual monthly check of prompts.

Where it runs

Online, as an article you can read in your browser on the creativeformore.com website.

Cost

Cost not stated on the official site.

Getting started

1. Open the article and first read the table that explains what each tool measures.
2. Identify the type of data you need: data supplied by the platform, or data gathered from test prompts.
3. Check prices and supported countries on each tool's official site before choosing.

What it is

Globatics is a web analytics service for websites that shows every visit on a globe that updates in real time. You turn it on by adding a script to your site; if you connect a payment provider, it also shows the sales linked to visits. It reports pages read, purchase paths, clicks that do not work, errors and alerts, and shows how many visitors come from ChatGPT, Perplexity, Claude and other assistants. By default it does not use cookies and does not store IP addresses.

Who it is for

It is useful for anyone who runs a website, blog or online shop and wants to know whether people arrive, read and buy. For example, someone who sells an online course can see how many people come from a Google search or from ChatGPT, and on which page they leave the purchase.

Where it runs

Online: the service is used from the browser, and the script goes into the site you want to measure. With the Pro and Business plans you can also connect an AI assistant, such as Claude, ChatGPT or Cursor, through the MCP server.

Difficulty

Easy for basic use, because you only copy a script into your site, a step the page says takes about two minutes. Connecting an AI assistant through MCP requires more technical familiarity.

Cost

A 5-day free trial with every feature, on one site and with no card. Then Starter is 9 € a month or 90 € a year, Pro is 29 € a month or 290 € a year, and Business is 99 € a month or 990 € a year. There is also a lifetime Pro licence for 299 € once, limited to 100 licences. Prices are in euros, VAT excluded; Polar adds VAT where it applies at checkout. If you do not choose a plan when the trial ends, the account is locked.

Getting started

1. Create an account for the 5-day free trial from the sign-up page, with no credit card.
2. Add the site you want to measure in the dashboard.
3. Copy the script into your site's code and use the install check to confirm it works.
4. If you want to see sales, connect Stripe, Polar or Lemon Squeezy.

What it is

StepMetrics is an online service that puts Google Search Console data side by side with the visits recorded on your site. It points to the page to fix first, for example one that Google shows many times but almost nobody clicks, and tells you what to try. After a change, it shows clicks from Google before and after, without claiming the result certainly comes from that change. It makes no promises about rankings and doesn't edit your site.

Who it is for

Useful for people who run a website, blog, or online store and want to see which pages perform well. For example, someone with a guide page that often appears in Google results but gets few clicks can see right away which title to rewrite.

Where it runs

Online: you use it from a browser, on the official site, with an account. Tracking works through a script you add to your site's pages. Programming assistants can connect via MCP. Installing it on your own server isn't available: every plan runs on the hosted service.

Difficulty

Medium: you have to add a script to your site's pages and connect Google Search Console. Anyone who doesn't work with websites can ask the site owner to handle these steps.

Cost

Free: $0 forever, 200,000 events per month, 2 sites, data kept for 400 days. Pro: $19 per month or $190 per year, 1,000,000 events per month, up to 100 sites, with clickstream. Scale: $99 per month, 10,000,000 events per month, up to 1,000 sites. Founding 50: $149 as a one-time payment for two years of Pro, for the first 50 buyers, with no automatic renewal. After two years the account moves to the Free plan if you don't subscribe. The service is subscription-based, with no open source license.

Getting started

1. Create a free account on the official site and add your site.
2. Add the tracking script to your pages, either yourself or by asking your programming assistant.
3. Connect Google Search Console from the Search page of the dashboard.
4. Open the report, read the suggested page, and note the changes you make so you can compare clicks later.

What it is

PresalesIQ is online software for managing the work of presales teams, meaning the technical staff who help sales reps close complex deals. After each call, it uses artificial intelligence to pull a structured record from transcripts, notes, and calendars, which the team reviews before saving. It then shows deal status, at-risk product tests, and product issues tied to revenue, and it can send them to Jira and the CRM.

Who it is for

It's useful for sales engineers, solutions consultants, and managers of technical sales teams who currently track deals in spreadsheets and chats. Example: a manager who wants to know which deals are at risk of falling through without having to ask each team member.

Where it runs

Online, from your browser, by logging in at app.presalesiq.ai. It connects to tools the team already uses, such as the CRM, calendar, and call recording.

Difficulty

Medium: you need to connect the CRM, calendar, and call recording tool, but the site doesn't describe any programming steps.

Cost

Cost not stated on the official site. The pricing page says the cost depends on the number of people on the team, optional modules, and integrations, and that the price is obtained in a short call. The base plan includes Engagements, Activities, and Demo Planning, while POC, Product Gaps, and AI Analysis are optional modules. No free plan or free trial appears.

Getting started

1. Book a demo or the pricing call from the official page at www.presalesiq.ai/book-demo
2. Tell them how many people will use the tool and which modules you need
3. Choose the integrations to connect from day one, such as CRM and calendar
4. Log in with your credentials at app.presalesiq.ai

What it is

ArtCraft is an open source studio for creating images and videos with artificial intelligence. You prepare the scene in 2D or 3D, place characters and objects, and set the camera. Then a model chosen by the user draws the result. The site cites models such as Seedance 2.5 and Nano Banana 2, and returns the generated images and videos.

Who it is for

Useful for people who work with images or video and want to decide framing, pose, and composition before generating. Example: the same character in several shots of the same room, without the objects changing from one scene to the next.

Where it runs

Online: in the browser, with nothing to install, at app.getartcraft.com. On your computer: a desktop app for Windows 10 (64 bit) or later and macOS 12.0 or later, with 8 GB of RAM recommended and 2 GB of free space. Linux does not appear on the download page.

Difficulty

Easy: installation takes three steps, and the browser version requires no installation. Building a well-made scene still takes some practice.

Cost

The program is free to download and use, under the Apache-2.0 or MIT open source license. Generating images and videos requires credits, or your own keys and subscriptions. Official plans (prices in dollars, with 20% off shown as a limited-time offer): Basic $8 per month instead of $10 per month, or $96 per year billed once a year, with 1,000 credits per month. Pro $28 per month instead of $35 per month, or $336 per year, with 3,750 credits per month. Max $48 per month instead of $60 per month, or $576 per year, with 6,600 credits per month. Enterprise on request. The page does not say whether taxes are included in the prices.

Getting started

1. Open the download page at getartcraft.com and download the file for Windows or macOS, or use the web version with no installation.
2. Install the program and sign in, or create a free account.
3. Create a scene: upload an image, remove the background if needed, and place characters or objects.
4. Generate images or videos with the model you prefer. Online generation requires credits, or you can use your own subscriptions or keys.

What it is

PigenAI is a website that generates images and videos with artificial intelligence. It gives access to more than 50 models from the same page, including Flux, Kling, Veo 3.1, and Ideogram. It includes text-to-image generation, AutoReel, which turns a script into a faceless video for YouTube and Shorts, and Effects, with 32 ready-made styles to apply to an uploaded photo. There is also motion control, which moves a character using a reference video.

Who it is for

It is useful for people who make social media content without wanting to show their face, or for people who want to try a visual style without writing complex instructions. Example: someone with a script for a short YouTube video who wants images and voiceover already edited together, or someone who wants to turn their own photo into a comic or anime style.

Where it runs

Online, directly in the browser, with nothing to install. Developers can also access it through an API, with REST endpoints compatible with OpenAI SDKs.

Difficulty

Easy: you use it in the browser, the effects don't require writing prompts, and creation starts from a plain-language description. Only using the API requires technical skills.

Cost

Free plan: 20 images and 3 videos per day, resetting every 24 hours, with no credit card required. Monthly plans: Starter $1.99/month for 150 credits, Premium $4.99/month for 400 credits, Creator Pro $9.99/month for 900 credits. One-time top-ups: $2.00 for 125 credits that do not expire. Taxes are not stated on the site. The site does not declare an open source license.

Getting started

1. Open https://www.pigenai.art/create and create a free account.
2. Write a description of the image you want and choose a model; the free plan has a daily limit.
3. For a video, try AutoReel by pasting your script, or use Effects by uploading a photo.
4. If you need more credits, check the pricing page before choosing a plan.

What it is

Vidnix is an online platform that generates videos and images with artificial intelligence. It brings several video models from other companies into one space, including Runway, Kling, Wan, and Seedance. It accepts a text, a script, or a photo and returns a video or an image. The site also describes workflows for transforming or extending existing videos.

Who it is for

Useful for creators who publish on YouTube Shorts, TikTok, or Instagram Reels, and for small businesses that want a promotional video. For example, someone with a product photo who wants to turn it into a short social media video, without a traditional production.

Where it runs

Online, from a browser, at vidnix.ai. The site lists no requirements for computers or servers.

Difficulty

Easy. The site says no editing experience is needed, since you start from a prompt or an image. Still, AI video takes several tries, and the site warns that a single generation is rarely enough.

Cost

On first access you get 30 free credits, enough to generate a few images. Generating videos requires purchased credits. The site shows a starting price of $0.20 per video with the Seedance 1.5 Pro model (the page says "as low as"). Individual plan prices are not readable on the official pricing page: Cost not stated on the official site. The site lists a 50% discount on the pricing page. Commercial use of the videos depends on the plan and the terms of service.

Getting started

1. Open vidnix.ai/it and choose Create Video or Create Image.
2. Sign up or log in: on first access you receive 30 credits, valid for images.
3. To generate videos, buy credits from the official pricing page (vidnix.ai/price).
4. Write a simple prompt with subject, action, setting, and style, generate the video, and refine the text if the result isn't convincing.

What it is

Crafting Apps is a collection of thirteen open source creative programs developed by the ArtCraft team and written in Rust. The apps cover image editing, vector illustration, video editing, photography, PDFs, animated graphics, page layout, interface design, word processing, audio, presentations, CAD technical drawing, and spreadsheets. They install on your computer and work with files stored locally.

Who it is for

Useful for people who want free tools to work on photos, PDFs, or presentations without subscriptions. For example, someone who needs to merge or split a PDF can look at PdfCraft, while someone looking for a text editor can look at WordCraft, which is still in development.

Where it runs

On a computer running macOS, Windows, or Linux, using installers. Twelve of the thirteen apps can also run in a browser, since the page notes WebAssembly support. No servers are required.

Difficulty

Medium: several apps are in alpha or still in development, so some features may be missing or may change.

Cost

Free. The apps page says the apps are “free to use” and open source. In the GitHub repositories for PhotoCraft, VectorCraft, FigCraft, WordCraft, PdfCraft, FilmCraft, EffectCraft, and LightCraft, the license is Apache-2.0.

Getting started

1. Open https://getartcraft.com/apps and choose the app you need.
2. Download the ArtCraft Launcher from the GitHub release page: it installs all the apps and keeps them updated.
3. Alternatively, download the installer for a single app from its page, selecting your operating system.
4. Check the app's status shown on the page: if it is in alpha or in development, try it first for testing and not for important work.

What it is

Keyflame is a site that collects free software, plugins, scripts, and resources for people who work with video and graphics. It has separate sections for Windows and Mac, written guides, video tutorials, and a list of new releases, kept up to date by a Discord community. It also has a section called Piracy, with an index of torrent sites and repacked versions of programs from third parties. The site says its downloads are verified, but it advises checking every file with an antivirus before installing it.

Who it is for

It is useful for people who edit video and want a guide to installing a plugin, for example for After Effects, without spending money. Anyone who does not want to risk files downloaded from unofficial sources should use only the websites of the program makers.

Where it runs

Online, from any browser, on any computer. Downloaded programs must be installed on a Windows or Mac computer.

Difficulty

Medium: the site is easy to browse, but downloaded files must be checked before installation, which takes a bit of care.

Cost

Free. The site says it is nonprofit and free to use. No pricing page is listed.

Getting started

1. Go to keyfla.me and choose the Windows or MacOS section.
2. Read the Beginner's Guide for basic settings and plugins.
3. Before installing any file, scan it with an antivirus and check the reports.
4. If you want to hear about news, join the Discord linked on the site.

What it is

Studio Zero edits raw video and turns it into a finished video on your own computer. The official home page says the program is in beta and includes 19 caption styles. The site does not say who makes it or list other features.

Who it is for

It may suit people who record videos and want quick editing with captions, for example someone preparing a video for social media who does not want to edit it by hand. The site does not describe other uses.

Where it runs

On your own laptop, as the home page says. The site does not list supported operating systems or minimum requirements.

Difficulty

Not stated on the official site. The page does not explain how to use the program or what is needed to install it.

Cost

Price not stated on the official site. The page lists no plans, prices, or license.

Getting started

1. Open the official site at studiozero.zilaxis.com.
2. Join the waitlist shown on the home page.
3. Wait for beta access. The site does not say when it will be available.

What it is

yt-dlp is a free, open source command-line program that downloads video and audio from thousands of sites. You start it with a command and give it the address of the page you want the content from. It saves a file on your computer, and it can also choose the format, download subtitles, and extract the audio.

Who it is for

It is useful for anyone who wants to keep a lecture, an interview, or a video published online on their computer, so they can watch it without an internet connection. For example, a teacher who saves a video to show in class on a day without network access.

Where it runs

On your computer, with supported systems: Windows, macOS, and Linux. You can use it as a ready-to-run executable file or install it with Python 3.10 or later. You also need ffmpeg to merge video and audio, and, for YouTube, a JavaScript runtime such as deno.

Difficulty

Medium: the installation is guided, but you use the terminal and must install ffmpeg separately and, for YouTube, a JavaScript runtime.

Cost

Free. It is open source software under the Unlicense license. The official page does not list paid plans.

Getting started

1. Go to the releases page on GitHub and download the file for your system, for example yt-dlp.exe on Windows or yt-dlp_macos on Mac. Alternatively, install Python 3.10 or later and use pip.
2. Install ffmpeg, which is used to merge video and audio and for later processing.
3. Open the terminal and type yt-dlp --version to check that the program responds.
4. Copy the address of the page you are interested in, then type yt-dlp followed by the link in the terminal.

What it is

CinderReels is a web social network inspired by the Cinderella fairy tale. It lets you post photos, videos, and text, and chat privately or in groups. Every post disappears at midnight in the time zone of the person who posted it. The official site does not appear to have a permanent archive of posts.

Who it is for

Good for people who want to share moments of their day with close friends without them staying online forever. For example, a private group of friends that trades photos from an evening and finds them gone the next day.

Where it runs

Online, from a computer or phone browser. The site says you can add it to your phone's Home screen without installing an app from the app stores.

Difficulty

Easy, since it works like a regular social network: just open the site, sign up, and post.

Cost

Price not stated on the official site.

Getting started

1. Open https://cinderreels.com in your phone or computer browser.
2. Create an account and complete your profile.
3. Post a photo, a video, or text, and choose whether to send it to your feed or to a group.
4. Check the countdown on the post: at midnight the content disappears.

What it is

Immich is an open source program that installs on your own server and collects photos and videos. It accepts files uploaded from your phone with the app or from a browser, and displays them in albums, maps, and searches by face, object, or date. It has a web interface and apps for iPhone and Android.

Who it is for

It is useful for people who have lots of photos on their phone and want to keep them under their own control. Example: a family that automatically saves everyone's phone photos to a server at home and shares them in common albums.

Where it runs

On your own server, using Docker, with at least 6 GB of RAM and 2 CPU cores according to the official quick start guide. Phone apps can be downloaded from the App Store, Google Play, the project website, or F-Droid.

Difficulty

Medium: installing the server requires Docker and a few terminal commands, while everyday use of the app is simple.

Cost

Free. The code is open source under the GNU AGPL v3 license. The official site has a 'Buy Immich' page to support the project, but no prices appear on the page read.

Getting started

1. Prepare a server or computer with Docker, at least 6 GB of RAM, and 2 CPU cores.
2. Download the docker-compose.yml and .env files from the official quick start guide, then set up the folder for photos and a database password.
3. Start the service with the command docker compose up -d from the folder you created.
4. Open the server address on port 2283 in your browser, create the first user (who will be the administrator), and then connect the phone app.

What it is

Cathode is a guitar program made by Rikarin, inspired by a professional effects processor. You connect it to your audio interface, then build your own sound by placing amps, cabinets, drives and effects on a grid, with scenes, channels and controls you can change live. It takes the guitar signal as input and outputs the processed sound, and it also loads NAM amp files and impulse responses (.wav, .cabgrid). The current version is a release candidate, so it is not stable yet.

Who it is for

It is useful for electric guitar players who want to try amp and pedal sounds without owning all the hardware. For example, a guitarist can test different amps and cabinets on a laptop before choosing which one to buy.

Where it runs

On your computer: macOS with Apple Silicon, Windows 10 and 11, and Linux desktops. Full audio support is on macOS only. Windows and Linux audio is still early. On a server: a small Linux computer such as a Raspberry Pi 5 or a mini PC, with no display, with your audio interface connected. You control it from a browser on a phone, tablet or laptop.

Difficulty

Medium. You need to connect an audio interface and set the audio device and buffer size. No programming is needed.

Cost

Free. The software is open source under the GPL-3.0 or later license. The official site says there are no accounts and no paid plans.

Getting started

1. Download the latest version from the Releases page on GitHub, choosing the package for your system.
2. Connect your guitar to your audio interface and open Cathode.
3. Go to Settings, then Audio setup, and choose your interface, sample rate and buffer size.
4. Double-click an empty cell on the grid and add an Amp and a Cab between Input and Output.

What it is

Mimi is an open source program released under the MIT license that runs on Mac, Windows, and Linux. It listens to the audio your computer plays, such as a movie or a lecture, and turns it into subtitles that stay on top of the screen. Speech recognition and translation go through a cloud service you choose, which receives the audio and requires your own credentials. It can show the original text, the translation, or both.

Who it is for

It is useful for people watching content in a language they don't know well who want to read the text while the video plays. For example, someone following an English lecture on a website who reads the translated subtitles without pausing the video.

Where it runs

On your computer: macOS 13 or later (Apple or Intel chip), Windows x64, and Linux x86_64 in preview (Ubuntu 22.04 or later). Recognition and translation require an internet connection and a cloud service. There is also an Android version, which must be installed from the source code and requires Android 10 or later.

Difficulty

Medium: the app installs like any other, but you must create an account with a cloud service, enter its credentials in the settings, and grant audio recording permission on macOS.

Cost

The program is free and open source under the MIT license. The cloud service for recognition and translation is paid by the user and may charge based on usage. Price not stated on the official site for individual providers. The project's README says Google Gemini offers a free plan, to be checked in Google AI Studio.

Getting started

1. Download the file for your system from the GitHub Releases page.
2. Open Settings, then Speech & Translation, add a configuration, enter the credentials for your chosen service, and save.
3. Choose the recognition language and the translation language.
4. Start a video and turn on Live Subtitles in the Subtitles section. On macOS, allow system audio recording when prompted.

What it is

Spectracraft is the first program from ResetLab Craft, a series of tools for DJs and producers that run on your computer. It analyzes tracks in MP3, WAV, AIFF, FLAC, AAC and ALAC by looking at the spectrum of the entire track, rates each file, summarizes whole folders, and saves reports in JSON. It works locally: it does not upload music anywhere and does not change files, tags or cover art. It is a public beta, not yet a stable release.

Who it is for

Useful to a DJ or producer who wants to know which files in their library have lower audio quality. Example: checking a folder of tracks before a gig and discarding those with a low rating.

Where it runs

On your computer: Windows x64, or a Mac with Apple silicon (M1 or later) and macOS 15 or later. There is not yet a package for Intel Macs or Linux.

Difficulty

Medium: you download a ZIP from the release page. The apps are not signed, so on Windows you need to unblock the ZIP before extracting it, and on the Mac you need to allow it to open in Privacy & Security.

Cost

Free and open source, GNU GPLv3 license.

Getting started

1. Open the v0.1.0-beta.1 release page on GitHub and download the ZIP for Windows or Mac.
2. On Windows: right-click the ZIP, choose Properties, then Unblock; then extract it into a new folder and open Spectracraft.exe.
3. On Mac: extract with Finder and open Spectracraft.app; if macOS blocks it, allow it in System Settings, Privacy & Security.
4. Add a track or a folder of tracks and start the analysis.

What it is

The Refiner is a web app that improves the audio of tracks generated with Suno and Udio, which often sound compressed. The site is run by Tiny Red Pixel Ltd. It accepts mp3, wav, or flac files up to 8 minutes long and returns a 44.1 kHz WAV file, or a 48 kHz file on the Pro and Studio plans. Beyond enhancement, it offers analog color control, mastering you can compare against the original, and stem separation, which splits a track into vocals, drums, bass, and the rest, available on the Pro plan. For now, the Flow Matching model and album processing are not available.

Who it is for

Useful for people who make music with Suno or Udio and want a cleaner file to upload to a distribution platform. For example, a songwriter who generated a song and hears it as dull or compressed can try the Enhanced option and compare the result with the original track before downloading it.

Where it runs

Online, directly in the browser, through the site therefiner.app. No computer installs are listed, and no options run on the company's own servers.

Difficulty

Easy: upload the track, leave the default Enhanced option selected, and listen to the result. Advanced options, such as the layer mixer, take a few more steps.

Cost

Free plan: 5 credits per month, $0. Hobby: $9/mo with 30 credits per month. Pro: $29/mo with 75 credits per month. Studio: $49/mo with 150 credits per month. One-time packs: 10 credits for $5 and 25 credits for $9, with credits that do not expire. Enhanced costs 2 credits per track. Mastering, AI artifact removal, and advanced mode cost 1 credit each. New verified accounts receive 5 welcome credits and 30 days of Pro features. It is not open-source software.

Getting started

1. Open therefiner.app and click "Start free" to create an account, no credit card required.
2. Upload a track in mp3, wav, or flac format, up to 8 minutes long.
3. Choose Enhanced, the recommended option, and start processing.
4. Listen to the result, compare it with the original, and download the WAV file you prefer.

What it is

HoloMapper is a real-time projection mapping program, meaning it projects images onto real surfaces such as walls, objects, or rooms. HoloMapper develops it and sells it on Steam. It accepts images, videos, GIFs, text, drawn shapes, and live sources through NDI or Spout, and sends the final output to a projector or a second screen. The site says the program is in Early Access, so some features may change.

Who it is for

It is useful for people who create visuals for events or concerts, people who prepare art installations, and people who want to try projection at home. For example, an artist can play a video on the front of a box or on the walls of a small room, adjusting the corners and surface points until the image lines up.

Where it runs

On a computer, with supported systems: 64-bit Windows 10 or 11, or Linux and SteamOS, including Steam Deck. Minimum requirements: Intel Core i3-2100 processor or equivalent, 4 GB of RAM, integrated graphics, DirectX 11 on Windows, and 200 MB of space. Recommended requirements: 16 GB of RAM, NVIDIA GTX 1050 or AMD RX 560 graphics card, and 2 GB of space. Output requires a projector or another screen. On Steam Deck, for projector output, the site says Desktop Mode with Proton Experimental may be needed.

Difficulty

Medium, because installation is simple but mapping requires understanding surfaces, layers, and warping (warp). The site says the program is designed for a direct workflow, so the learning curve is manageable.

Cost

$29.99 on the official Steam page, with a one-time payment. The site says there are no subscriptions: you buy it and own it. The Steam page also says the price could gradually rise during Early Access as new features arrive. No free version or free trial appears on the official pages reviewed.

Getting started

1. Open the HoloMapper page on Steam and buy it at the listed price.
2. Install it on a Windows or Linux computer, or on a Steam Deck.
3. Connect a projector or a second screen, and set up clean output from the program menu.
4. Read the official documentation on surfaces, layers, and mesh warping before starting your first project.

What it is

LearnHouse is an open source platform for creating and managing online courses. Users write lessons in a block editor, add videos, documents, and code, and assign homework with submissions. It also includes forums, podcasts, completion certificates, and an AI assistant for students and teachers. Payments go through your own Stripe account.

Who it is for

Useful for teachers, schools, corporate trainers, and creators who want their own space for courses with a members-only area. Example: a language teacher uploads video lessons, exercises to submit, and a final certificate for anyone who completes the course.

Where it runs

Online, with the cloud service on the official site. Or on your own computer or server with self-hosting: Docker and Node.js are required, and it runs on Linux, on macOS with Docker Desktop, or on Windows through WSL2. Minimum server requirements: 2 GB of RAM, 10 GB of storage, 1 core, and ports 80 or 443 free.

Difficulty

Medium. The online service starts with sign-up, but installing the version on your own server requires Docker and terminal commands.

Cost

Free plan, with limits that the official pages describe differently. Paid plans listed on the official site: Standard at $49 per month and Pro at $149 per month, with an annual option that saves 20% (annual figures not visible in the text read). Personal plan at $15 per month for people who take courses. The code is licensed under AGPL-3.0. Enterprise features have a separate license and price on request.

Getting started

1. Go to learnhouse.app and create an account for the online service, or read the installation guide if you want to use it on your own server.
2. For self-hosting, check that Docker and Node.js are installed, then run the command npx learnhouse@latest setup in the terminal: the guided setup asks for domain, database, and administrator account.
3. Create the first organization and the first course from the admin panel.
4. Check the plan's member and administrator limits before inviting many learners.

What it is

Maiths is an independent site that collects past IGCSE math papers, both CAIE and Edexcel. The student completes the paper in the browser, saves answers, and can ask AI for help on a single question. When finished, they can download the paper, have it graded by AI, or compare it with the marking scheme. The site states that it does not provide an official result and that important decisions should be checked against official material and with a teacher.

Who it is for

Useful for students preparing for the IGCSE math exam who want to practice with real papers, and for teachers who need to review the work done. Example: a student completes the June 2023 Foundation paper, code 4MA1, then asks for help on one geometry question before having it graded.

Where it runs

Online: the paper is completed directly in the browser, as the site's page indicates. The site does not say which browsers or systems are supported.

Difficulty

Easy: the work is done in the browser, but you need to choose the right paper carefully (code 4MA1, Foundation or Higher tier, session and year).

Cost

Free plan: $0 with the free limits listed on the site (the pricing page does not list them). Basic Student: $6.66 for grading only or $9.99 with all features. Pro Student: $13.33 for grading only or $19.99 with all features. Pro Teacher: $16.66 for grading only or $24.99 with all features. Ajiro Tech: free with an eligible access code, not available for purchase. Billing period and VAT: not stated on the pricing page.

Getting started

1. Open the Past papers page and choose the exact paper: code 4MA1, year, session, tier (Foundation or Higher), and paper number.
2. Complete the paper online and save your answers. For individual questions, you can ask AI for help next to the related work.
3. When the paper is finished, choose whether to download it, have it graded by AI, or compare it with the marking scheme.
4. Check the pricing page to see what is included in the Free plan before you pay.

What it is

AppFlowy is a workspace for writing documents, keeping wikis, and managing projects and databases. The project presents itself on GitHub as an open source alternative to Notion, and the official site describes it as an AI workspace that can run on your own infrastructure. The AI features include search with source citations, automatic meeting notes, and automatic filling of columns. It does not convert files: you work directly inside the app, and the company says it offers help transferring data from Notion and Confluence.

Who it is for

It is useful for small teams that want an internal wiki and a project list without entrusting their data to an external service. Concrete example: a group that keeps its task calendar, meeting notes, and standard operating procedures in one place, and wants to sync them between computers and phones.

Where it runs

Online, in the browser. On computers with macOS, Windows, or Linux. On iPhone and Android. Or on your own server (self-hosting), following the official guide; the minimum server requirements are not listed in the pages I read and should be checked in the guide.

Difficulty

Easy for normal use, since you only need to open the browser or download the app. You need a technical person only if you want to install the version on your own server, because the official guide is written to reach a production environment.

Cost

Personal plan: $0 per month, with up to 100 MB of storage and files up to 7 MB. Pro plan: $12 per member per month with annual billing, or $16 per member per month with monthly billing. The self-hosting price does not appear on the pricing page I read, so I am not listing it. The software is open source under the AGPL-3.0 license.

Getting started

1. Open https://appflowy.com/app and create a free account, or download the app from the download page on the official site.
2. Create your personal workspace and start with a first page or a database.
3. If you work in a team, consider the Pro plan from the settings, in the Billing section.
4. To install it on your own server, follow the official self-hosting guide from the first page to the end.

What it is

UTD is a column-based task board where you drag cards from one column to another. It is built to work alongside Claude Code, which can read tasks, move them and comment on them through a server included in the project. Every change is recorded with the name of who made it and appears in real time on all open browsers.

Who it is for

It is useful for people who use Claude Code and want to see how far the work the assistant is doing has got. For example, a developer puts the jobs in an "Up Next" column, asks Claude Code to work through them in order, and follows the cards as they move on the board.

Where it runs

On your computer, locally, with Python 3.12 or later and Node 22 or later. Or on a server, using the project's Docker image. The project warns against exposing it to the internet without protection, because it does not have a sign-in yet.

Difficulty

A technical person is needed: setup happens in the terminal, with commands, environment variables and Docker, and the program does not have a login yet.

Cost

Free: the project is released under the MIT license (open source). No prices are listed on the official site. The cost of an optional server is not stated.

Getting started

1. Open the project's GitHub page and read the README file.
2. Install Python 3.12 or later with uv, and Node 22 or later.
3. In the project folder, run "make migrate && make seed && make dev", then open http://localhost:5173.
4. Open the project folder with Claude Code: the .mcp.json file registers the utd server.

What it is

FilePizza is a website that lets you send files from one device to another through your browser. Alex Kern and Neeraj Baid created it, and it is an open source project on GitHub. The sender picks a file and gets a link to send to the recipient; the file goes straight to the recipient's browser and does not pass through the site's servers.

Who it is for

It is useful for anyone who needs to send a large file to someone without using an upload service. Example: sending a family video to a relative in another city while both keep the browser open.

Where it runs

Online, right in the browser, including browsers on phones. The project can also be installed on your own server with Docker, following the instructions on GitHub. The site does not list minimum requirements.

Difficulty

Easy to use from the website, since you only choose the file and share the link. Installing it on your own server requires a technician, because it involves terminal commands.

Cost

Price not listed on the official site. The code is released under the BSD 3-Clause license. The site invites visitors to support development with a donation.

Getting started

1. Open https://file.pizza in the browser on your computer or phone.
2. Drag the file onto the page or select it from your device.
3. If you want, set an optional password to protect the transfer.
4. Send the link to the recipient and keep the window open until the transfer is complete.

What it is

OneBeep is a Mac app that lives in the menu bar at the top of the screen. It shows every app that is playing sound with its own volume slider, so you can turn down Spotify without touching the ongoing call. From the same panel, you can also change the device where the audio comes out, without opening System Settings.

Who it is for

It's useful for people who use a Mac with several programs making noise at once, such as a phone call or a meeting while music plays in the background. With a single movement, you can lower the music and keep the call's voice loud.

Where it runs

On your computer, running macOS 14.4 or later. The official page lists only Mac.

Difficulty

Easy, since it installs like a normal app, and the site states that no drivers or restart are needed.

Cost

Free. The site lists “Free” and “No subscription,” meaning no subscription fee. To support development, there is an optional tip in USDT, a cryptocurrency, with a wallet address on the page. The price of any other versions is not stated on the official site.

Getting started

1. Check that your Mac has macOS 14.4 or later.
2. Go to onebeep.tech and click “Download OneBeep.”
3. Launch the app: the icon appears in the menu bar at the top.
4. Start playing music or a video, open the panel, and drag the slider of the app you want to lower.

What it is

PasteDrop is a website for moving text, links, and files from one device to another. You create a room, open it on the other device with the code or QR code, and whatever you type or paste appears on both at once. Images, PDFs, and ZIP files up to 10 MB pass directly between devices. Rooms are temporary and are deleted after 5 minutes of inactivity; the site says it does not keep data.

Who it is for

Useful for anyone who needs to quickly copy something from a phone to a computer, or the reverse, without emailing it to themselves. Example: a link opened on the phone that you paste into the browser on the computer.

Where it runs

Online, in a browser, on both phone and computer. No app to install and no sign-up needed.

Difficulty

Easy: open the site, create a room, and scan the QR code with the other device.

Cost

Cost not stated on the official site. No sign-up needed.

Getting started

1. Open pastedrop.xyz and choose Create Room.
2. On the other device, scan the QR code or enter the room code.
3. Type, paste, or drag in a file: it appears on the other device right away.

What it is

Logimu is a service that collects information on products sold on Amazon and Walmart: price, ranking position, customer rating, and all the sellers offering the same item. Someone searching for a product gets a sorted list of results. People who work with data can also read price and availability history. It accepts a keyword, an ASIN code, a barcode, or an Amazon link, and returns the data in structured format.

Who it is for

Useful for online sellers who want to check competitors' prices, for people looking for product ideas, or for anyone using an AI assistant to compare items. Example: asking Claude which coffee machine costs less than 80 dollars with high ratings, and getting a list with the date each price was recorded.

Where it runs

Online. Use it at logimu.com, through the internet API, or as a connector in Claude, ChatGPT, and other compatible apps by adding https://api.logimu.com/mcp. The Claude connector requires a paid Claude plan.

Difficulty

Medium: the assistant connector takes a few steps to set up, but using the API directly requires some programming skills.

Cost

Free plan with no sign-up: 30 requests a day through the assistant connector. Paid plans, prices in dollars as listed on the official pricing page: Solo at $19 a month, Starter at $50 a month, Growth at $125 a month, Standard at $250 a month, Scale at $1,000 a month, Volume at $3,000 a month. The Solo plan includes 25,000 credits a month. Refund available within 14 days, and you can cancel at any time. It is not open-source software.

Getting started

1. Open the free demo at logimu.com and try a search, no account needed.
2. To use it in Claude or ChatGPT, add https://api.logimu.com/mcp as a connector. No key is needed.
3. To use the API or run broader searches, create a key on the site by choosing a paid plan.
4. Always check the recording date shown next to the data.

What it is

An open-source metasearch engine: it sends the same query to several engines (Google, Bing, DuckDuckGo, Brave, Qwant and others) and shows the results on a single page. It does not track or profile users.

Who it is for

People who want to search without leaving a profile with search engines. And people who build AI assistants or agents and want to connect them to a search engine they run themselves.

Where it runs

Online, through public instances run by volunteers: as of 15/10/2026, searx.space listed 71 active ones. Open the address and search. Or on your own server, with the container or the official installation script.

Difficulty

Easy with a public instance. Running your own requires a technician.

Cost

Free, open source under the AGPL-3.0 license. On your own server you only pay for the server.

Getting started

1. Open searx.space and choose an instance with high uptime and low response times.
2. Set it as your browser's search engine, if you like it.
3. For your own instance, follow the installation guide at docs.searxng.org: the recommended method is the container.

What it is

Patina is a website that keeps a record for each car: technical specs, installed parts, costs, receipts, and photos. It lets you search for replacement parts by part number or in plain words, and opens online stores with a tap. When you sell a car, it generates a link with the full repair history, including the date, mileage, and cost of each job.

Who it is for

It's useful for people who do their own repairs, people with several cars to maintain, and mechanics or car dealers. Example: someone who changes brake pads can note the €70 replacement part, the €120 labor, and the receipt in the car's record, so they know how much they've spent so far.

Where it runs

Online, from your browser. It works well on a phone too, and nothing needs to be installed.

Difficulty

Easy, because nothing needs to be installed and you log in with a link sent by email, with no password.

Cost

Free plan at €0, for up to two cars, with all features included. Pro plan at €15 per month or €150 per year (two months free), for unlimited cars. The site says the prices include VAT. Cars on accounts created before Pro launched stay unlimited for free.

Getting started

1. Open patinaservice.com and enter your email in the field «Try it with your email».
2. Open the access link that arrives by email. No password is needed.
3. Add your first car with make, model, engine, license plate, mileage, and photo. If you don't have a photo, the site looks for one.
4. Record parts, costs, and receipts as you go. To search for parts without signing up, use the «Find a part» page.

What it is

PlainlyOS is an online service that reads a hard-to-understand document, such as a bill, a letter, or a contract. It explains what it means in everyday language, highlights dates, amounts, and things that may need attention, and suggests next steps. It accepts PDF, Word, and JPG or PNG image files.

Who it is for

It can help anyone who receives a renewal notice, a craftsman's estimate, a medical bill, or a condo association communication and wants to know what it asks. Example: someone who gets an automatic renewal of a service and wants to know the deadline for canceling. The site states that it does not provide legal, medical, or financial advice.

Where it runs

Online, from your browser, after logging in to an account. The site also mentions a version you can use as a web app on your phone.

Difficulty

Easy, because you only need to upload the document and read the explanation. No technical skills or program installation required.

Cost

Each account starts with 3 free analyses. After that, the site lists $6.99/ month. You can cancel at any time, and cancellation takes effect at the end of the current billing period. Payment is handled through Stripe.

Getting started

1. Go to https://plainlyos.com and click «Get started» to create or open your account.
2. Before uploading a file, remove card numbers, passwords, personal identification numbers, and health data, as the site recommends.
3. Upload a bill, letter, or contract in PDF, Word, JPG, or PNG format.
4. Read the explanation, dates, amounts, and suggested steps, then ask follow-up questions about the document.

What it is

CityTour AI is a phone app that builds walking routes for a city. You choose the city, the time you have, and the themes you care about, and the AI builds the route with stops, a map, and stories. At each stop you find an audio guide, photos, and sources, and you can ask the AI questions.

Who it is for

It's useful for people visiting a city alone or as a couple who want to walk at their own pace, without a group or fixed schedule. For example, someone with two free hours in Rome who loves architecture can ask for a 90-minute route on that theme.

Where it runs

On your phone, as an app for Android (Google Play) and iPhone (App Store). You need the app to create a tour. Tours can be prepared for offline use, so you can follow them even when the signal is weak.

Difficulty

Easy, because it installs like any normal app and you use it by choosing city, length, and interests, with no technical settings.

Cost

The app is free and includes a full Berlin tour. Other tours, both ready-made and generated, are unlocked with credits bought in the app as single packs, with no subscription. Credit prices are not stated on the official site: you see them in the app before buying, and they depend on the country and the store.

Getting started

1. Download the app from Google Play or the App Store.
2. Open the free Berlin tour to see how it works.
3. Choose the city, the length (45, 90, or 180 minutes), and your interests, then generate your route or open a ready-made tour.
4. Prepare the tour for offline use before you head out, if the signal outside might drop.

What it is

Vela Maps is a free, open source maps app for Android, designed as an alternative to Google Maps. The map comes from OpenStreetMap and other open data. Real-time traffic, opening hours, reviews, and place photos are requested from Google anonymously only when you search for something, with no account and no Google services on the phone. It also works offline if you download the region in advance, and its voice guidance runs on the phone. It is in beta and is developed by a single person.

Who it is for

It is useful for people who use an Android phone without Google services, for example with GrapheneOS, and still want traffic and store hours. For example, someone heading to an area with poor coverage can download the region before leaving and use the map, search, and navigation without a connection.

Where it runs

On the phone, as an app for Android 8 or later. Install it with Obtainium, from the project's F-Droid repository, or by downloading the APK file from the releases page.

Difficulty

Medium: the app works like Google Maps, but it must be installed with Obtainium, F-Droid, or an APK file.

Cost

Free, with no ads and no account. Licensed under GPLv3; the developer writes that they will never ask for money for the app.

Getting started

1. Open velamaps.org and go to the Install section.
2. Choose how to install it: Obtainium, the project's F-Droid repository, or the APK file from the releases page.
3. Before a trip, download the region or country you need so the app works without a connection.

What it is

Online service that connects travelers with people who live in Korea, for short, personal experiences. The site presents it as an alternative to the organized tour: hours without a script spent in someone's everyday life. The listed office is in Seoul.

Who it is for

Useful for visitors to Korea who want to see a neighborhood the way people who live there see it, for example taking a stroll with a local and stopping to eat at a spot in the area.

Where it runs

Online, at the website wanderat.org.

Cost

Cost not stated on the official site (pricing page not verified).

Getting started

1. Open wanderat.org and read how the experiences work.
2. Go to the sign-in page to create a profile or log into your account.
3. Look for the experience you're interested in and read what it includes before booking.
4. For questions, write to hello@wanderat.org, the address listed on the site.

What it is

ease.fm is a free music streaming site with instrumental tracks made by an AI music model. You don't need to sign up. Press Start Quickplay and a queue of tracks begins. Giving a track a thumbs up or down reorders the queue around the genres you prefer.

Who it is for

Useful for anyone who wants background music with no lyrics and no ads, for example someone studying in the evening or working at a computer who wants to stay focused.

Where it runs

Online, in any modern browser on phones, tablets, and computers. Nothing to install.

Difficulty

Easy, because you press one button and the music starts. No account or setup is needed.

Cost

Free. The site says every track and playlist is free to stream. No paid plans appear on the pages reviewed.

Getting started

1. Open ease.fm and press Start Quickplay.
2. Give a thumbs up to tracks you like and a thumbs down to ones you don't.
3. To choose genres, use Filter by genre or start from Synthwave, Lo-fi, Game Music, or Rock.
4. To keep a selection, save it as a playlist. It stays in the browser you are using.

What it is

AIPass is a free, open source program from the AIOSAI organization that adds to a project folder. It creates AI agents that keep the memory of earlier sessions and send messages and tasks to each other. It works with Claude Code: AIPass manages the agents and Claude Code does the work.

Who it is for

It is useful for people who build an app or website with AI help and want to pick up the work the next day without starting over. Example: someone working on a project folder who wants the agent to remember the decisions made the evening before.

Where it runs

On your computer, from the terminal. The official site lists Linux or macOS, and Windows through WSL2. The GitHub page also lists Windows with Git Bash, but notes that WSL is untested. You need Python 3.10 or later and Claude Code.

Difficulty

Needs a technical user: it runs from the terminal, you must install Python 3.10 or later and Claude Code, and the installer changes some computer settings, such as Claude Code rules and your shell startup file.

Cost

Free: AIPass is open source under the MIT license. To run it you need a Claude subscription that includes Claude Code. The official site lists the Pro, Max, Team or Enterprise plans. The GitHub page mentions only Pro or Max. The price of the Claude subscription is not listed on the official AIPass site.

Getting started

1. Check that you have Python 3.10 or later, Claude Code, and a computer with Linux or macOS (on Windows, use WSL2).
2. Download the project from GitHub with git clone https://github.com/AIOSAI/AIPass.git, enter the AIPass folder, and run ./aipass install.
3. Or install with pipx install aipass, then run aipass install, as the official site describes.
4. Enter your project folder and run aipass init run. The next time, just say hi to your agent to pick up where you left off.

What it is

Dessa is a speech model built by Block's AI research lab. It takes audio in and sends audio out, and it can listen while it talks, so it can break in and add short reactions during a conversation. Block says it was trained only on recordings of human conversation, with no text.

Who it is for

It is aimed mainly at researchers in voice AI and AI safety, the group Block is giving early access. A regular user can for now only listen to the sample recordings on the site to hear what a conversation with the model sounds like.

Where it runs

Online, to listen to the sample recordings on the official site. Block says the model runs locally on a MacBook Pro M4 Max or better, but the site does not offer a download yet.

Difficulty

Needs a technician: there is no public download yet, and running it takes a very high-end computer.

Cost

Price not listed on the official site. The page shows no plans, rates, or license.

Getting started

1. Open block.xyz/dessa and listen to the sample recordings.
2. Check the Download section: the site says launch links are coming soon.
3. If you research AI safety, email ragavan+dessa@block.xyz to ask for early access.

What it is

CLSTR is a site that gathers articles from many news outlets and groups them by event, so the same story appears only once. The World page lists, for each country, the stories covered most in the last three hours, with the number of sources reporting them. Each story links to a page that follows how the situation develops.

Who it is for

For people who want to know what is happening in the world without opening ten news sites. For example, someone following an international crisis or an election who wants to see which outlets are covering it.

Where it runs

Online, in the browser, at clstr.news. The World page opens without signing in.

Difficulty

Easy, because it reads like a news site and needs no setup.

Cost

Free plan at $0: 3 searches per day, 7-day history window, up to 2 saved searches, 1-country filter. Pro plan: $179 per year, plus taxes, or $14.92 /mo billed annually. The monthly price of Pro is not shown in the text of the pricing page. For the API plans (Free $0, Builder $29/mo, Scale $299/mo) taxes are calculated at checkout.

Getting started

1. Open https://clstr.news/world in your browser.
2. Scroll the list of countries and choose the story you care about.
3. Open the story page to see updates and the sources covering it.
4. If you want emails when a story changes, consider the Pro plan on the pricing page.

What it is

CLSTR is an online news service. It groups articles from many outlets that cover the same event into one event, and links related events into stories that can last weeks. The Observatory page shows statistics on 164 days of news, starting April 10, 2026, and is frozen at September 20, 2026. CLSTR reports what sources say and does not check whether they are right.

Who it is for

Useful for anyone who wants to see how a crisis or topic develops without reading the same story on ten different sites. For example, someone who wants to know how an international dispute changed over the last few weeks.

Where it runs

Online, in a web browser, at clstr.news. For scheduled access (API or MCP server) nothing needs to be installed: a free key and an internet connection are enough.

Difficulty

Easy to read the page and use the site. The API needs technical skills, because you send requests to web addresses with a key.

Cost

The Observatory page is free. Free plan for people: $0, with 3 searches a day and a 7-day history window. Pro plan: $179 per year, plus taxes, or $14.92 /mo billed annually. API: Free $0; Builder $29/mo; Scale $299/mo, billed monthly. Caps are hard limits with no overage charges.

Getting started

1. Open the Observatory page and read the figures, noting that the data is frozen at September 20, 2026.
2. Use "Browse live situations" to see the stories currently on the site.
3. Create a free account to get the email briefing three times a week (Monday, Wednesday and Friday on the Free plan).
4. For developers: create a free API key in Settings and try the example request on the API page.