> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentkey.app/llms.txt
> Use this file to discover all available pages before exploring further.

# GitHub data for AI agents

> GitHub repositories, READMEs, contributors, issues and pull requests, issue search, developer profiles and activity, and trending repos for AI agents.

GitHub is where open-source software is built in public: who maintains a
project, how active it is, what users keep reporting and who is rising are all
there, spread across repository pages, issue threads, profiles and the Trending
page. GitHub's own API asks for a token and rations requests per hour, and
Trending has no official API at all. Through AgentKey your agent reads public
GitHub data the same way it reads everything else — `find_tools`,
`describe_tool`, `execute_tool` — starting from a repository URL or a
username.

## What you can ask

* *"What is this repository — stars, language, license, and when was it last updated?"*
* *"Read the README of this repo and summarize what it does and how to install it."*
* *"Who are the top contributors to llama.cpp, and where do they work?"*
* *"List the open issues on this repository updated since 1 October, most commented first."*
* *"Find merged pull requests labelled 'bug' in openai/whisper this year."*
* *"Which forks of this project have the most stars?"*
* *"Show this developer's profile, their repositories and the pull requests they opened in the last three months."*
* *"GitHub 本周最火的 Python 项目有哪些？"*

## What's covered

**Repositories**

* Metadata by repository URL — stars, forks, primary language, topics,
  license and timestamps
* The README of the default branch, as the Markdown the maintainers wrote
* Contributors ordered by commit count, with each one's login and profile link
* Forks sorted by newest, oldest or most starred, with each fork's owner,
  description, stars, language and last push

**Issues and pull requests**

* A repository's issues and pull requests, filtered by state (open, closed or
  all) and update date and sorted by creation, last update or comment count —
  with author, state, labels, counts and body
* The comments on one issue or pull request, oldest first, with each
  commenter's relationship to the repository and the reactions
* Search across GitHub with GitHub's own qualifiers — `repo:`, `is:pr`,
  `is:issue`, `author:`, `label:`, `created:` and the rest — sorted by best
  match, comments, reactions, creation or last update

**Developers**

* Public profile by username — name, bio, company, location, followers and
  repository counts
* A developer's public repositories — name, description, language, stars and
  forks — sorted by creation, last update, last push or name
* Followers and the accounts a developer follows, page by page
* The pull requests a developer has opened, within an optional date range
* Contribution activity for a year — monthly summaries of commits, pull
  requests and issues with repository links — and the contribution graph with
  per-day counts

**Trending**

* Trending repositories — rank, stars, stars gained, language and
  description — for today, this week or this month, optionally for one
  language
* Trending developers with their most popular repository, by language and
  time range

A typical task chains these: start from Trending or a search, open the
repository and its README, list its contributors, then profile the ones that
matter.

## How to use it

Pass the request in plain language and let the router pick the tool:

```text theme={null}
find_tools(q="top contributors to this GitHub repository")
find_tools(q="GitHub 本周最火的 Python 项目")
find_tools(prefix="developer/github")      # browse everything for GitHub
describe_tool("<returned Provider/Operation>")
execute_tool("<Provider/Operation>", { ...params })
```

Repository tools take a URL such as `https://github.com/owner/repo`, the
comment tool takes an issue or pull request URL, and developer tools take a
username — no numeric ids. List tools page by number or by cursor; keep the
page size the same while walking pages. Always `describe_tool` before
executing — it shows each tool's filters, sort options and per-call cost.

> [!NOTE]
>
> **Building with code?** GitHub is available through the catalog tools above
> only; there are no fixed-contract Scenario Methods for it yet. See the
> [Scenario Methods overview](https://docs.agentkey.app/api-reference/scenario/overview.md) for the
> platforms that have them.

## FAQ

> [accordion] Do I need a GitHub token?
>
> No. AgentKey holds the upstream access; your agent only needs your AgentKey
> subscription.

> [accordion] Can it read private repositories or act on GitHub?
>
> No. Only public repositories, issues, pull requests and profiles are
> available, and every tool is read-only — no issues opened, comments posted
> or stars given.

> [accordion] Can it read source files?
>
> The README is covered; other files are not. For a research paper's code
> repository, alphaXiv's `read_files_from_github_repository` in
> [Academic](https://docs.agentkey.app/capabilities/academic.md) opens files directly, and
> [Scrape](https://docs.agentkey.app/capabilities/scrape.md) can fetch any other public file URL as text.

> [accordion] How fresh is the data?
>
> Every call goes to GitHub at request time; nothing is served from a stale
> archive. Trending reflects what GitHub's Trending page shows at that
> moment.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.