Import skills from GitHub
Load any GitHub user or organization, auto-detect AI repositories — MCP servers, agents, prompt libraries, skills, plugins, workflows — and install them into your AI Brain.
Enter a GitHub user or org (try , , or ) and hit Load repos.
What this page is for
A great deal of the useful work in AI right now is published as a GitHub repository rather than as a product. MCP servers that connect a model to a database or an API, agent frameworks, prompt collections, skill definitions, automation workflows — most of it arrives as a repo with a README and no install page.
Finding it is the hard part. GitHub search returns everything and ranks by stars, so a repository that does exactly what you need but was published last month is effectively invisible. This page takes the opposite approach: name an organisation or a person whose work you already trust, and see what of theirs is relevant.
How the detection works
Enter a GitHub user or organisation and we read their public repositories through GitHub's own API, then classify each one by signals in its name, description, topics and language — an mcp- prefix, an agent topic, a prompt library structure, and so on. What comes back is categorised rather than listed, so you can see at a glance whether an organisation publishes MCP servers, agent tooling, or neither.
The classification is heuristic and it is wrong sometimes. A repository with a vague name and no topics may be miscategorised or missed entirely — the signal it is reading is the metadata the author chose to publish, and not everyone fills it in. Treat the categories as a starting point rather than an authoritative index.
Before you install anything
This is worth reading rather than skipping. An MCP server or agent tool runs with whatever access you grant it, and by design that is often broad — your files, your API keys, your database. A malicious or simply careless one is a real risk, and the fact that a repository has stars is not a security review.
Read the source before you run it, particularly anything that requests credentials or network access. Prefer repositories with recent commits and a visible maintainer over abandoned ones with more stars. Where you can, run it with the narrowest permissions that make it useful rather than the broadest that make it convenient.
Saving a repository to your AI Brain here only bookmarks it — it does not execute anything, and nothing is installed on your machine by this page. Installation happens on your own system, on your own terms.
Where to start
If you are new to this, the model providers' own organisations are the safest first look — they publish reference implementations that are worth reading even when you do not need them. Our guide to MCP explains what those servers actually do, and the Skills Index maps the wider landscape without requiring you to browse repositories at all.