github-repo-mcp
Provides tools for analyzing public GitHub repositories, including repo health scoring, recent commit summaries, good first issue discovery, and semantic Q&A over README and docs.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@github-repo-mcp Check the health of facebook/react and find a good first issue."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
github-repo-mcp
An MCP (Model Context Protocol) server that lets any AI agent — Claude Desktop, Claude Code, Cursor — reason about a public GitHub repository directly, instead of the human manually reading READMEs, commit logs, and issue trackers.
Why this exists
Evaluating an unfamiliar repo ("is this actively maintained? is this issue actually beginner-friendly? what changed recently?") is repetitive manual work every developer does. This server turns that into four callable tools any MCP-compatible agent can use.
Related MCP server: GitHub MCP Server
Tools
Tool | What it does |
| Deterministic 0–100 health score from commit recency, issue-triage ratio, and contributor count. Flags archived repos immediately. |
| Groups commits from the last N days by type (feature/fix/docs/refactor/chore) for a quick "what changed" view. |
| Finds beginner-labeled issues, scores each for genuine clarity, and flags ones that look mislabeled (e.g. touches core architecture despite the "good first issue" tag). |
| Free-text Q&A over README + |
Setup
npm install
cp .env.example .env
# Add a GitHub personal access token to .env — raises the rate limit from 60/hr to 5000/hr.
# No special scopes needed for public repo data: https://github.com/settings/tokens
#
# ask_repo also needs, both free tier:
# VOYAGE_API_KEY — https://dashboard.voyageai.com/ (API Keys -> Create new key)
# QDRANT_URL, QDRANT_API_KEY — https://cloud.qdrant.io/ (create a free cluster)
npm run buildRunning locally with Claude Desktop
Add this to your Claude Desktop MCP config (claude_desktop_config.json):
{
"mcpServers": {
"github-repo-mcp": {
"command": "node",
"args": ["/absolute/path/to/github-repo-mcp/build/index.js"],
"env": { "GITHUB_TOKEN": "your_token_here" }
}
}
}Restart Claude Desktop, then try: "Use github-repo-mcp to check the health of facebook/react and find me a good first issue."
Architecture notes
GitHub API, not scraping — clean, documented, generously rate-limited with a token.
Deterministic scoring, generative explanation —
get_repo_health's score is a fixed formula (recency + issue triage + contributor diversity), not an LLM guess. This keeps the number reproducible; an LLM (in the calling agent) can narrate why on top of it.Mislabel detection in
find_good_first_issues— real "good first issue" labels are often wrong. The tool checks body length, red-flag keywords (architecture, migration, security), and comment count to catch issues that are mislabeled, not just present them at face value.
Demo
See docs/demo.md for a real (not mocked) protocol exchange against
facebook/react, including the embeddings-based ask_repo retrieving the correct
answer for a query that shares no words with the source text.
How ask_repo retrieval works
ask_repo started as keyword-window retrieval over the README (see git history) and
was upgraded to embeddings-based RAG, implemented in src/rag.ts:
On first query for a repo, chunk README +
docs/*.md(top-level, cap 15 files) + top 8 most-discussed issue threads (title + body, ~500-word chunks, 50-word overlap)Embed each chunk with Voyage AI (
voyage-3.5-lite)Store vectors in Qdrant Cloud, one collection per repo (
repo_<owner>_<name>)On every query: embed the question, cosine-similarity search top-5 chunks
Subsequent queries for the same repo skip re-indexing (checked via the collection's
points_count) — only the query itself gets embedded
Why embeddings over keyword matching: keyword overlap requires literal shared words — "how do I install React?" and a README section titled "Getting Started" share zero words and would never match. Embeddings place semantically similar text near each other in vector space regardless of exact wording, so retrieval survives paraphrasing and synonyms — the actual failure mode keyword search hits in practice.
Known simplifications (marked ponytail: in src/rag.ts), each with a stated
upgrade path:
Issue threads index title+body only, not the full comment discussion (avoids one extra API call per issue) — fetch
/issues/{n}/commentsif answers need to reflect resolution discussion, not just the original report.Chunk sizing uses word count as a token-count approximation, not a real tokenizer.
No staleness check — a repo indexed once stays indexed even if its docs change. Add a TTL or a
pushed_atcheck against the stored index if content goes stale.
License
MIT
Available Tools
4 toolsask_repoAsk RepoA
Answers a free-text question about a repo via embeddings-based semantic search over its README, docs/*.md, and top issue threads. Use this for questions the other structured tools can't answer directly, e.g. 'does this support TypeScript?' or 'how do I configure X?' Requires VOYAGE_API_KEY and QDRANT_URL/QDRANT_API_KEY to be set.
Inputs: repo (required), owner (required), query (required).
find_good_first_issuesFind Good First IssuesA
Finds open issues labeled for beginners/first-time contributors, then scores each for genuine clarity of scope and flags ones that look mislabeled (e.g. touches core architecture despite the 'good first issue' label).
Inputs: repo (required), owner (required).
get_repo_healthGet Repo HealthA
Computes a maintenance-health score (0-100) for a public GitHub repo from commit recency, issue-triage ratio, and contributor diversity. Use this before deciding whether to depend on or contribute to a repo.
Inputs: repo (required), owner (required).
summarize_recent_commitsSummarize Recent CommitsB
Fetches commits from the last N days and groups them by type (feature/fix/docs/refactor/chore) so an agent can describe what changed without reading raw commit logs.
Inputs: repo (required), owner (required), days.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 4 tools
Each tool targets a clearly distinct purpose: health scoring, commit summarization, beginner-issue discovery, and semantic Q&A. There is no meaningful overlap—ask_repo explicitly serves as a fallback for questions the structured tools cannot answer.
All tool names use lowercase snake_case and follow a predictable verb-first pattern: get_, summarize_, find_, ask_. The consistency makes it easy to infer each tool's action at a glance.
Four tools is a well-scoped count for a focused repository-analysis server. Each tool covers a distinct aspect of evaluating or exploring a repo, and none feel redundant or missing from the apparent purpose.
The server covers the core repo-evaluation workflow well: health, recent changes, contribution entry points, and open-ended questions. Minor gaps exist, such as structured access to license, stars, or PR activity, but ask_repo can work around many of these.
Maintenance
Related MCP Connectors
Code intelligence for LLMs. Analyze, search, and retrieve code from any public git repository.
GitHub repo search, stats, language breakdown and comparison as agent tools.
Connect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
Operate GitHub from your AI assistant: repos, PRs, issues, Actions. Previews and read-only mode.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables Large Language Models to analyze GitHub repositories in real-time, providing tools for retrieving repository information, analyzing issues, accessing documentation, and visualizing activity.-
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to interact with GitHub repositories, issues, pull requests, code, and more through a comprehensive set of tools.-
- AlicenseNot gradedqualityDmaintenanceEnables Claude to analyze GitHub repositories with tools for health scoring, contributor analysis, issue tracking, code search, and more.MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to perform GitHub analytics and management tasks, including repository analysis, PR summarization, issue triage, release notes generation, and contributor statistics.39 npmMIT