GitHub Org Intelligence (MCP)

What does GitHub Org Intelligence do?

Pass a GitHub org, repo, or user and get back a structured intelligence card: language mix, stars, activity level, top repos, top contributors, latest releases, recent commit cadence, and a deterministic LLM-ready markdown report. No LLM calls inside, fully reproducible. Designed to be called from an MCP agent loop (Claude Code, Cursor, Continue, custom agents) where the agent needs to evaluate a GitHub presence in one shot.

Built on the public GitHub REST API. Optional GitHub PAT (free, no scopes needed) lifts the rate limit from 60/h to 5000/h.

Why use it from an AI agent?

• One call, full picture - org metadata + top 15 repos + language mix + activity classification + LLM-ready report.
• Deterministic - same input today, same output tomorrow (no LLM jitter).
• Tight schema - agents don't waste tokens parsing the GitHub REST output.
• Activity-aware - the actor classifies dormant / maintained / active / very_active per repo and per org so the agent knows whether a project is alive.
• Standby-mode HTTP server - true MCP-over-HTTP endpoint with one-shot fallback.

Modes

A. Org Overview

Analyze an organization or user account: portfolio summary, language mix, top repos.

{
    "mode": "org_overview",
    "org": "anthropics",
    "topReposLimit": 15
}

B. Repo Intelligence

Single repository deep dive: language byte breakdown, top contributors, releases, recent commit frequency.

{
    "mode": "repo_intelligence",
    "repo": "anthropics/claude-code"
}

C. User Profile

Developer profile: account metadata + their public repos + topic interests.

{
    "mode": "user_profile",
    "username": "torvalds"
}

Output shape (all modes)

{
    "meta": {
        "mode": "org_overview",
        "query": "anthropics",
        "scrapedAt": "2026-05-08T15:00:00Z",
        "primaryLanguage": "TypeScript",
        "totalStars": 32450,
        "publicRepos": 47,
        "activityLevel": "very_active"
    },
    "intelligence": {
        "primaryLanguage": "TypeScript",
        "languageMix": [
            { "language": "TypeScript", "repoCount": 18, "percent": 38.3 }
        ],
        "totalStars": 32450,
        "publicRepos": 47,
        "activityLevel": "very_active",
        "activeRepos90d": 28,
        "archivedRepos": 5,
        "forks": 3,
        "medianStarsPerRepo": 47,
        "licensesUsed": ["MIT", "Apache-2.0"],
        "topRepos": [
            {
                "name": "claude-code",
                "fullName": "anthropics/claude-code",
                "url": "https://github.com/anthropics/claude-code",
                "stars": 12340,
                "language": "TypeScript",
                "activityLevel": "very_active"
            }
        ]
    },
    "org": {
        "login": "anthropics",
        "name": "Anthropic",
        "description": "...",
        "type": "Organization"
    },
    "llm_ready": {
        "summary": "Anthropic on GitHub: 47 public repos (28 active in 90 days), 32450 total stars, primarily TypeScript...",
        "markdown_report": "## GitHub Org Intelligence Report\n..."
    }
}

Authentication

The Actor calls GitHub's public REST API. Two auth modes:

Anonymous (default)

60 req/hour. Fine for low-volume agent use (1-2 calls / agent session). No setup required.

BYOK GitHub PAT (optional, recommended for higher volume)

Lift the rate limit to 5000/hour by supplying a Personal Access Token. No special scopes needed - even a no-scope PAT works for public data access. Two ways to provide it:

1. Per-call via githubToken input field (encrypted by Apify).
2. Operator-scoped via GITHUB_TOKEN env var on the Actor (Apify Console -> Settings -> Environment variables).

Generate a PAT at https://github.com/settings/tokens?type=beta (fine-grained, public-repo read).

Pricing (PPE)

Mode	Price per call
All modes	$0.001 actor start + $0.005 intelligence card = $0.006/call

Activity classification

Level	Criterion
very_active	Last push <= 7 days ago
active	Last push 8-30 days ago
maintained	Last push 31-180 days ago
dormant	Last push > 180 days OR archived

For org/user mode, the org-level level reflects the distribution across the org's repos (e.g. very_active if 30%+ of live repos pushed in the last week).

Calling from an MCP agent

The Apify MCP server (mcp.apify.com) exposes this Actor as the tool call-actor with name github-org-intelligence-mcp. Typical prompt: "Use github-org-intelligence-mcp to analyze the anthropics org's GitHub portfolio."

Direct Standby HTTP integration:

curl -X POST -H 'Content-Type: application/json' \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -d '{"mode":"repo_intelligence","repo":"vercel/next.js"}' \
  https://<actor-standby-url>

Limits

• 60 req/h anonymous, 5000/h with PAT. The actor consumes 1-3 GitHub API calls per intelligence card depending on mode.
• Repo language byte breakdown is GitHub's own classification (occasionally noisy on monorepos with vendored deps).
• Contributors list caps at 100 per page; very large repos may have more total contributors than reported.

Sources, freshness, legality

• All data comes from GitHub's public REST API. No HTML scraping, no auth required for public data.
• The Actor exposes only public GitHub content. Private repos / orgs / users not accessible.
• User-Agent identifies the Actor with a descriptive string per GitHub API best practices.