Git & GitHub CLI MCP Server (v0.4.1)

Industrial-grade Git and GitHub orchestration for the Agentic Revolution. Built for developers who need their AI implementation engines to interact with repositories without environment friction.

Stack: v0.4.1 FastMCP 3.2.0+ Python 3.12+ Windows 10/11 GitHub CLI (gh)

Industrial Disclosure: This server is part of the Agentic Revolution Manifesto. All tools are designed for machine-native consumption, providing high-fidelity telemetry and crash-resistant automation for the 135+ repository Alsergrund fleet.

---

🚀 Zero-Friction Orchestration

If your AI agent fails to perform Git or GitHub actions because of gh CLI path issues or authentication gaps, this server solves that.

• Automatic Discovery: It finds your gh.exe installation in standard Windows locations (Program Files, Scoop, Winget) without requiring a system PATH entry.
• 100+ Tools: Complete coverage for Git (local) and GitHub (remote) workflows.
• Industrial Strength: Zero "fluff"—just direct, actionable results for commits, PRs, issues, and repository management.
• SOTA v14.x Automated Startup: Root-level start.ps1 and start.bat for one-click deployment of the full FastAPI + Vite stack.

---

Capabilities

Git Operations (43 actions)

Group	Operations
CORE	init, clone, add, commit, push, pull, fetch, status
INSPECT	log, diff, show, blame
BRANCH	create, switch, delete, merge, rebase, list
REMOTE	list, add, remove
STASH	stash, pop, list, drop
TAG	list, create, delete
UNDO	reset, revert, cherry_pick
CLEANUP	clean
SUBMODULE	add, update, sync, status
BISECT	start, bad, good, reset
WORKTREE	add, list, remove

GitHub Operations (58 actions)

Group	Operations
REPOS	list, view, create, fork, clone, delete, rename, archive
ISSUES	list, view, create, close, comment
PRs	list, view, create, merge, checkout, close, comment
RELEASES	list, view, create, upload, delete
GISTS	list, view, create, edit, delete
WORKFLOWS	list, view, run, enable, disable
LABELS	list, create, edit, delete
SECRETS	list, set, delete
ORGS	list, view
USERS	user_view
SEARCH	repos, issues, prs, code, users
AUTH	status, login, logout

---

Design Principles

• Integrated Tooling: Consolidates multiple CLI utilities into a single, reliable server interface.
• Structured Feedback: Every tool returns a consistent result object for easy agent parsing.
• Low Friction: Minimal configuration required. If gh is installed, it works.

Maintainer: PR triage (small repos)

Silence reads as indifference. You do not need a bot on day one: use github_ops to list open PRs and post a short first reply.

1. List open PRs (includes comments and updatedAt so you can spot threads that sat cold):

   github_ops(operation="pr_list", owner="YOUR_USER", repo="YOUR_REPO", state="open", limit=50)

2. Acknowledge so the author knows it was seen — github_ops(operation="pr_comment", owner=..., repo=..., pr_number=..., body="..."). Example tone (edit to your voice):

   Thanks for the PR — I maintain this in spare time and do not always see notifications quickly. I have read it and will review properly within the next few days; I will comment here if I need changes.

3. GitHub.com: for repos you care about, set Watch to All activity (or at least Participating) so PRs are not only visible when you open the site.

4. Optional: add a PR template or CONTRIBUTING.md line setting expectations (e.g. “I try to acknowledge within 48h; deep review may take longer”).

5. Web inbox: with the FastAPI + Vite app running, open /inbox — Pull requests & Issues with optional fleet mode (many owner/repo lines), stale highlights, and the same data as github_ops. A supervisor stack (OpenClaw, OpenManus, RoboFang, OpenClaude, etc.) can run a daily heartbeat that calls pr_list / issue_list across your repo list; this page is the human-facing mirror.

6. Fleet doc (mcp-central-docs): GITHUB_MAINTAINER_HEARTBEAT.md — where the schedule lives, robofang council hook, prompt sketch.

This does not replace a fair review — it reduces “I wasted hours in the void” as the default story.

🛡️ Industrial Quality Stack

This project adheres to SOTA 14.1 industrial standards for high-fidelity agentic orchestration:

• Python (Core): Ruff for linting and formatting. Zero-tolerance for print statements in core handlers (T201).
• Webapp (UI): Biome for sub-millisecond linting. Strict noConsoleLog enforcement.
• Protocol Compliance: Hardened stdout/stderr isolation to ensure crash-resistant JSON-RPC communication.
• Automation: Justfile recipes for all fleet operations (just lint, just fix, just dev).
• Security: Automated audits via bandit and safety.

License

MIT