# GitHub > Canonical: https://www.yalc.ai/mcps/github/ Read repos, manage issues, and trigger Yalc skill installs from inside Claude Code. ## Categories - operations **Website:** https://github.com/ **Repo:** https://github.com/github/github-mcp-server ## Glance - Maintainer: GitHub (official) - License: MIT - Auth: GitHub PAT or OAuth **Install Command:** claude mcp add github --env GITHUB_PERSONAL_ACCESS_TOKEN=ghp_xxx -- docker run -i --rm -e GITHUB_PERSONAL_ACCESS_TOKEN ghcr.io/github/github-mcp-server **Install Notes:** GitHub ships an official MCP server distributed as a Docker container. Generate a Personal Access Token at github.com/settings/tokens with the scopes you need (repo, read:org, read:user, write:discussion). Replace `ghp_xxx`, run the command, restart Claude Code. There is also an npm-based community alternative for environments without Docker. **Yalc Fit Score:** 8 **Yalc Verdict:** The skill repo and issue tracker layer for Yalc operators. Pull skill source files, manage GTM-OS issues, and respond to PRs without leaving Claude Code. ## Plain Description GitHub maintains an official MCP server (`github/github-mcp-server`) that exposes the GitHub API as native Claude tool calls. Coverage spans repositories, issues, pull requests, discussions, code search, file reads, gists, and user profiles. Distribution is via Docker container, which is the most stable option for production agents. For Yalc operators, GitHub is the skill repo backbone. Cloning skills, reading SKILL.md files, opening issues to request new skills, and reviewing skill PRs all happen here. The MCP makes those operations one-shot prompts instead of CLI context switches. ## Yalc Framework **Workflow Step:** route ### Workflow Narrative The GitHub MCP sits at the **route** node when Yalc workflows need to coordinate skill development or look up skill source code. It pairs naturally with the Yalc GitHub repo (the open-source skill set). Most useful patterns: surface SKILL.md content for a specific skill on demand, open issues to request skill features, batch review weekly PRs, and pull weekly contribution stats for the Yalc skill set. **Workflow Position:** The code and skill repository node. Yalc reads SKILL.md files, opens issues, and writes PR descriptions. Less about CRM, more about the Yalc skill ecosystem itself. ### Prompt Patterns - Yalc, search the YALC GitHub repo for skills tagged "linkedin". Read each SKILL.md and produce a one-paragraph capability summary per skill. → Yalc uses GitHub MCP to search and read files, summarizes via Claude. - Yalc, open a new issue on the YALC repo titled "Skill request: HubSpot CRM writeback". Body: 3 paragraphs from this Notion brief. → Yalc creates the issue with structured body via the MCP. - Yalc, list every open PR on YALC, classify as "ready for review", "blocked", or "draft", and post a Slack digest to #engineering. → Yalc reads PR list, classifies, posts via Slack MCP. ### Chaining **Upstream:** Yalc prompt → GitHub MCP (read or write) **Downstream:** GitHub event (issue, PR) → Yalc workflow (via webhook handler) ### Anti Patterns - Don't grant the broadest PAT scope (admin:org, delete_repo). Use the minimum scopes needed. If the MCP errors with permissions, scope up incrementally. - Don't use the MCP to clone whole repos. The MCP reads files; for clone operations use `gh repo clone` directly. Wrong tool for the job. - Don't read large files (above 1MB) through the MCP. Token usage explodes. Use targeted file reads or grep. **Compatibility:** The official server runs in a Docker container, which works in Claude Code, Claude Desktop, and Cursor when Docker is installed. Community npm versions exist for Docker-less environments. GitHub's API rate limits apply (5000 req/hour for authenticated PATs). ## Pros - Official GitHub-maintained server. MIT licensed. - Covers repos, issues, PRs, discussions, code search, file content. - PAT auth simpler than OAuth for personal use. - Docker container is production stable, no Node version drama. - Free to use. GitHub API rate limits are generous. ## Cons - Docker requirement. Friction in environments without Docker installed. - PAT scopes can be confusing. Wrong scope means cryptic permissions errors. - File read responses can be huge. Use targeted paths, not whole-repo reads. - GitHub Actions workflow inspection is shallow. For deep CI inspection use the gh CLI directly. ## Who For - Yalc operators who maintain or extend the open-source skill set - Engineering managers tracking PR throughput and issue triage - Anyone whose Claude Code workflow includes "look up the source for skill X" ## Related ### Skill **Rule:** Pair when you need to scrape GitHub pages outside the API surface. ## Alternatives - **GitLab MCP** - Rule: Switch when your repo lives on GitLab instead of GitHub. - Url: # - **Direct gh CLI** - Rule: Skip the MCP for batch operations. The gh CLI is more efficient for headless scripts. - Url: # - **Composio GitHub toolkit** - Rule: Switch when you want unified credential management across multiple Composio toolkits. - Url: # ## Faq - Q: Do I need Docker to use the GitHub MCP? - A: For the official server, yes. There are community npm alternatives if Docker is a non-starter, but the official Docker version is the most stable. - Q: What PAT scopes does the MCP need? - A: For typical Yalc workflows you need repo (private repo access), read:org (team and org reads), read:user (user info). For issue and PR writes, the repo scope covers it. For discussions, add write:discussion. - Q: Can the MCP create commits? - A: Yes via the contents API (create or update file). For multi-file commits, the create-tree API is exposed. For complex commit operations, gh CLI is usually simpler. - Q: Is the MCP rate limited? - A: GitHub's API allows 5000 authenticated requests per hour. Typical Yalc volume (a few hundred reads per day) is nowhere close. Watch for rate limits when batch-reading files across many repos. - Q: Can I use the MCP with GitHub Enterprise? - A: The official server supports GitHub Enterprise via the `--gh-host` flag. Check the repo README for configuration details. - Q: Does the MCP handle GitHub Apps differently than PATs? - A: PATs are simpler for personal use. GitHub App auth (with installation tokens) is the right path for production agents acting on behalf of an org. The MCP supports both; pick based on your auth model. **Reviewer:** Othmane Khadri