MCP Server

Repo Crawler for AI agents.

Turn any GitHub repo into structured intelligence. Metadata, issues, security alerts, SBOMs — all through one tool call.

Get started Read the Handbook

Quick scan

get_repo_summary({ owner: "facebook", repo: "react" })

Deep crawl

crawl_repo({ owner: "myorg", repo: "api", tier: "3" })

Compare

compare_repos({ repos: [{ owner: "vitejs", repo: "vite" }, ...] })

Features

Everything an AI agent needs to understand a repository.

5 MCP Tools

Crawl repos, crawl orgs, summarize, compare, and export — each purpose-built for a different workflow.

3-Tier Data Model

Start light with metadata, go deeper with issues and PRs, or get the full security picture with Dependabot, SBOM, and secret scanning.

Section-Selective

Only calls the APIs you ask for. Request specific sections to save quota and shrink response size.

Graceful Degradation

A 403 on Dependabot won't kill the crawl. Each section fetches independently with permission tracking.

Rate-Limit Safe

Built-in Octokit throttling with automatic retry on 429s. No more blowing through your API quota.

Safe Exports

CSV with formula injection prevention, Markdown with pipe escaping. Export crawl data without security risks.

Quick Start

Add to your MCP client config and start crawling.

Claude Code / Claude Desktop

{
  "mcpServers": {
    "repo-crawler": {
      "command": "npx",
      "args": ["-y", "@mcptoolshop/repo-crawler-mcp"],
      "env": {
        "GITHUB_TOKEN": "ghp_your_token"
      }
    }
  }
}

Try it

// Quick repo triage (4 API calls)
get_repo_summary({ owner: "expressjs", repo: "express" })

// Full security audit
crawl_repo({ owner: "myorg", repo: "api", tier: "3" })

// Compare frameworks side-by-side
compare_repos({ repos: [
  { owner: "sveltejs", repo: "svelte" },
  { owner: "vuejs", repo: "core" },
  { owner: "facebook", repo: "react" }
]})

Data Tiers

Choose the depth of data you need. Each tier includes everything below it.

Tier

What you get

API calls

Tier 1 — Fundamentals

Metadata, file tree, languages, README, commits, contributors, branches, tags, releases, community, workflows

~11

Tier 2 — Activity

Tier 1 + issues, PRs, traffic, milestones

~15+

Tier 3 — Security

Tier 2 + Dependabot alerts, security advisories, SBOM, code scanning, secret scanning

~20+

Tools

Five MCP tools covering the full GitHub data surface.

`crawl_repo({ owner, repo, tier, sections? })`

Crawl a single repository at any data tier. Supports section-selective fetching and configurable limits for commits, issues, PRs, and alerts.

`crawl_org({ org, tier, min_stars?, language? })`

Crawl every repo in an organization with filters for stars, language, forks, and archived status.

`get_repo_summary({ owner, repo })`

Quick human-readable summary using only 4 API calls. Ideal for triage and initial assessment.

`compare_repos({ repos, aspects? })`

Side-by-side comparison of 2–5 repos. Stars, languages, activity, community health, and size.

`export_data({ data, format, sections? })`

Export crawl results as JSON, CSV, or Markdown. CSV includes formula injection prevention.