github-mcp-proxy
A self-hosted GitHub MCP server with per-repository guardrails for AI agents, enabling fine-grained access control and audit trails.
README
github-mcp-proxy
A self-hosted GitHub MCP server running on Cloudflare Workers, designed for personal / small-team use with per-repository guardrails for AI agents.
Any MCP-capable agent (Claude, Codex, ChatGPT, Cursor, Cline, …) can be plugged in — with tight control over which repos it can touch and how.
[!WARNING] If you need security or stability, use the official GitHub MCP server. Fork freely; security issues are your responsibility once forked.
Why
Generic GitHub tokens give an agent full account access. This server narrows that down:
- Per-repo permission levels —
read/issues/full - Branch protection — new branches must match a prefix (default
claude/);main/masternever writable - Commit-message policy — Conventional Commits, validated server-side
- Label whitelist —
allowedLabelsper repo; the server never auto-creates labels on GitHub - Forbidden paths —
.github/workflows/**always blocked - Audit trail — every write auto-appends
Signed-off-by/Assisted-bytrailers or a footer identifying the agent and model
Architecture
MCP agent ──(Streamable HTTP + OAuth)──▶ Cloudflare Worker
│
├── workers-oauth-provider (OAuth 2.1 server to MCP clients)
├── GitHub OAuth (OAuth client to github.com)
├── KV (session tokens)
├── Durable Object (MCP agent state)
└── Octokit (GitHub REST)
Config (ghmcp.config.ts) is bundled with the Worker at deploy time — no runtime KV reads, typechecked via satisfies.
Quick start
git clone <your-fork>
cd kz-github-mcp
npm install
cp ghmcp.config.sample.ts ghmcp.config.ts # edit: repos you want to expose
cp wrangler.sample.jsonc wrangler.jsonc # edit: KV id + worker name
cp .env.example .env # edit: for local dev only
Create a GitHub OAuth App (Homepage = https://<worker>.<subdomain>.workers.dev, Callback = /callback), then:
wrangler kv namespace create "OAUTH_KV" # add the id to wrangler.jsonc
wrangler secret put GITHUB_CLIENT_ID
wrangler secret put GITHUB_CLIENT_SECRET
wrangler secret put COOKIE_ENCRYPTION_KEY # openssl rand -hex 32
npm run deploy
Local development: same flow with a second OAuth App pointing at http://localhost:8788, credentials in .env, then npm run dev.
Continuous deployment (optional)
.github/workflows/deploy.yml deploys to Cloudflare on every push to main. Set two repo secrets in GitHub Settings → Secrets and variables → Actions:
| Secret | Where to get it |
|---|---|
CLOUDFLARE_API_TOKEN |
Cloudflare dashboard → My Profile → API Tokens → create a token with the "Edit Cloudflare Workers" template |
CLOUDFLARE_ACCOUNT_ID |
Cloudflare dashboard → right sidebar of any Workers page |
The GitHub OAuth credentials (GITHUB_CLIENT_ID, GITHUB_CLIENT_SECRET, COOKIE_ENCRYPTION_KEY) live as Wrangler secrets on the Worker itself, not in GitHub Actions — you only need to set them once with wrangler secret put.
See docs/ARCHITECTURE.md for how it works, docs/CONFIG.md for every config field.
Connect an AI agent
Claude (claude.ai) — Settings → Connectors → Add custom connector → paste https://<worker>.<subdomain>.workers.dev/sse.
Claude Code / Codex / Cursor / Cline — add to the client's MCP config:
{
"mcpServers": {
"github": {
"command": "npx",
"args": ["mcp-remote", "https://<worker>.<subdomain>.workers.dev/sse"]
}
}
}
Any MCP host — point it at the same /sse endpoint. OAuth flow runs on first connect.
Tools
Prefixed with ghmcp_. Each declares a minimum permission level; the server denies calls below that.
| Category | Tools | Min level |
|---|---|---|
| User | get_me |
read |
| Repo | list_repos, get_file_content, search_code |
read |
| Issues | list_issues, get_issue, search_issues, list_issue_comments |
read |
| Issues | create_issue, update_issue, add_issue_comment |
issues |
| PRs | list_pull_requests, get_pull_request |
read |
| PRs | create_pull_request |
issues |
| Files | create_or_update_file, create_branch |
full |
| Actions | list_workflow_runs, get_workflow_run_logs |
read |
Any tool can be disabled via disabledTools in the config.
Validation
npm run validate-config
Uses the gh CLI to confirm that every repo in ghmcp.config.ts exists and that every label in allowedLabels is present on GitHub. Runs automatically as a prebuild hook, so a typo fails npm run build.
License
Originally forked from cloudflare/ai/demos/remote-mcp-github-oauth. MIT-licensed.
This tool is MIT-licensed. — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.