Craft Anything
A free, open-source Rust engine that studies any website, app or game with 314 inspection tools, then has a coding agent build and check your own version.
Maigret checks a username against 9,978 sites, pulls the profile data each one exposes and follows the new handles it finds. MIT licensed, no API keys.
Maigret is a free, open source OSINT tool that takes a username and checks whether an account with that name exists on 9,978 sites. It extracts the public details each profile page exposes, follows the new usernames and identifiers it finds, and writes the findings to HTML, PDF, JSON, CSV or a graph. The licence is MIT, and it needs no API keys.
You give it a username. Maigret loads its site database, asks each site whether the name is taken, and records the answer. A default run checks the 500 highest-ranking sites by traffic, which keeps a first search to a few minutes. Add -a and it checks the full database, which takes much longer. Use --tags photo,dating or a country tag such as --tags us to narrow a run to one slice of the list.
For every hit it can go further than "the account exists". Maigret uses socid_extractor to pull names, bios, profile links and internal IDs from the page and from the site's API, including links to accounts on other platforms. It then runs a second pass with any new username or ID it discovered, so one search can uncover accounts you did not know to look for. --parse URL starts a run from a profile page, and --permute builds likely variants from two or more inputs, so "john doe" also checks johndoe and j.doe.
A machine with Python 3.10 or newer, or the standalone Windows executable, the Snap package or Docker. There is no account to create and no key to request, and the tool runs from a command line, a local web interface or a Python import.
Network access matters more than hardware. Some sites block datacenter addresses or answer with a challenge page, so Maigret accepts an HTTP or SOCKS proxy, plus Tor and I2P gateways. The site database refreshes from GitHub once every 24 hours and falls back to the copy shipped with the package when you are offline.
| Item | Detail |
|---|---|
| Licence | MIT, checked in the repository's LICENSE file |
| Latest release | 0.6.6, published 18 September 2026 |
| GitHub stars | about 40,900, checked 19 November 2026 |
| Python | 3.10 or newer, below 4.0 |
| Sites in database | 9,978, of which 9,230 are enabled (database dated 17 November 2026) |
| Default scan | 500 highest-traffic sites; -a checks the rest |
| Input | usernames, one or several |
| API keys | none required |
| Output | HTML, PDF, XMind, JSON, CSV, text, D3 graph, Neo4j script |
| Platforms | Windows, macOS, Linux, Docker, cloud notebooks |
Coverage is the reason people pick it. Nine thousand sites is far past what a person can check by hand, and the database is tagged by site type and country, so a search can stay focused. The repository publishes sites.md with the database date, the enabled count and a breakdown of check strength, which makes the tool's own accuracy auditable rather than a promise.
The database is also maintained. The latest commit, from 18 November 2026, fixes nine false positives found by a full run against random usernames. That kind of upkeep is what keeps a username tool usable, because site rules and page layouts change constantly.
The extraction step is the other half of the value. A bare "account found" tells you little. Pulling the display name, bio, links and numeric IDs from a profile gives you material to correlate, and the recursive pass turns a single handle into a small map of accounts.
Output options suit different jobs. HTML for reading, PDF for a file you can send, XMind for a mind map, JSON and CSV for your own scripts, a Neo4j Cypher script for a graph database, and an interactive graph in the browser. The built-in web interface at --web 5000 shows the results graph and offers the same downloads from one page.
It also runs where you need it: pip install maigret, a standalone Windows executable, a Snap for Linux, two Docker images for CLI and web use, Colab and Binder notebooks, or import maigret inside a larger Python pipeline. A community Telegram bot covers the case where you cannot install anything at all.
It only takes usernames. Email addresses and phone numbers are outside its scope, so it answers a narrower question than "who is this person".
A hit is a lead, not an identification. People reuse handles, and plenty of them share common names. The project's own statistics say 4,511 of 9,230 enabled checks (48.87%) rely on weak signals, and 3,710 (40.2%) infer existence from the HTTP status code alone. Treat each result as something to verify against the profile, not a conclusion.
Blocking is a real constraint. Of the 9,978 sites, 246 sit behind bot protection, spread across TLS fingerprinting, Cloudflare JavaScript challenges, IP reputation and custom systems. The Cloudflare bypass needs a local FlareSolverr instance and is marked experimental, so its behaviour can change without notice.
The headline number is not the default run. Unless you pass -a, you are checking 500 sites, and a full sweep of 9,230 enabled sites takes far longer and meets more blocks along the way.
The web interface has no login out of the box. The README says so for the one-click Render deployment: anyone with the URL can use it. If you host it yourself, put authentication in front.
Finally, the legal side is yours. The tool is published for lawful, educational use, and the disclaimer puts GDPR, CCPA and local law on the person running the search. Looking up your own accounts, a brand you manage, or a subject you are authorised to investigate is the intended use.
The shortest path, on any system with Python 3.10 or newer:
pip install maigret
maigret YOUR_USERNAME
Save an HTML report alongside the console output:
maigret YOUR_USERNAME --html
Search the whole database, or narrow it by tag:
maigret YOUR_USERNAME -a
maigret YOUR_USERNAME --tags photo,dating
On Windows, download maigret_standalone.exe from the releases page, then double-click it or run it from Command Prompt with flags such as --html. No Python install is needed for that route.
With Docker, the CLI image writes reports to a mounted folder, and the web image starts the browser interface:
docker run -v /mydir:/app/reports soxoj/maigret:latest username --html
docker run -p 5000:5000 soxoj/maigret:web
On Linux, sudo snap install maigret installs a confined package with no Python dependency. For the local web interface from a normal install, run maigret --web 5000 and open http://127.0.0.1:5000.
PDF output is an optional extra: pip install 'maigret[pdf]', with system graphics libraries needed on Linux and macOS.
There is an optional AI pass, --ai, which sends the collected results to an OpenAI-compatible endpoint and streams a short summary. That is the only feature that needs a key, set as OPENAI_API_KEY or in settings.json.
Yes. It is MIT licensed, which allows commercial use without a fee. The author sells two separate services for teams that need them: a private site database updated daily, and a username check API. The open source tool does not need either.
No. The site checks run without credentials. Only the optional --ai analysis needs an OpenAI-compatible key and endpoint.
The database lists 9,978 sites as of 17 November 2026, with 9,230 enabled. A run without flags checks the 500 highest-traffic sites, and -a checks the rest.
No. It takes usernames. An email address or phone number is not a valid input for the site checks.
It means the site returned a profile page for that name, which is a lead. Roughly half the enabled checks use weak signals by the project's own count, and handles get reused, so confirm anything important against the profile itself.
The database falls back to the bundled copy when GitHub is unreachable, but the checks query live sites, so the tool needs a network connection to produce results.
Yes. The CLI wraps an async function you can import, which lets you run searches inside a script and filter sites by tag. The documentation covers the library pattern.
Maigret sits in our developer tools and software analysis section, next to the free utilities that inspect code and binaries. If you would rather run the web interface on your own server than from a laptop, the self-hosted section covers the tools that make that practical.
A free, open-source Rust engine that studies any website, app or game with 314 inspection tools, then has a coding agent build and check your own version.
LiteLLM is a self-hosted gateway giving every LLM provider one OpenAI-format endpoint, with virtual keys, budgets, spend tracking and fallbacks.
Jenkins runs build, test and deploy jobs from a Jenkinsfile kept in your repository, on your own hardware with no per-minute billing.