MCP.so
Sign In
P

Patchwork OS MCP Server

@Oolab-labs

Basic information

Category

Productivity

Transports

stdio

Author

Oolab-labs

Submitted by

Oolab

About Patchwork OS MCP Server

A local-first MCP bridge for Claude Code, Claude Desktop, and other MCP clients. 177 tools (LSP, debugger, terminal, git, GitHub, file ops) via a companion VS Code extension, with an optional layer for YAML automation recipes, a human approval queue, persistent decision traces, a

Patchwork OS MCP Server config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "patchwork-os": {
      "command": "npx",
      "args": [
        "-y",
        "patchwork-os"
      ]
    }
  }
}

Patchwork OS MCP Server tools

Tool list not available yet

This listing does not currently have a tool list to display. This does not mean the server has no tools; check its documentation for supported capabilities.

Fetch the live tool list by running this server in a temporary sandbox using the button above.

Overview

What is Patchwork OS?

Patchwork OS is a local-first personal AI runtime that runs on your machine, under your control. It combines an MCP bridge for Claude Code IDE integration with a full automation platform featuring YAML recipes, a delegation policy with an approval queue, durable trace memory, and OAuth—all with no vendor lock-in and no telemetry unless you opt in.

How to use Patchwork OS?

Install the npm package (npm install -g patchwork-os). For the Claude IDE Bridge run claude-ide-bridge and connect Claude Code with claude --ide. For the full Patchwork OS run patchwork init to scaffold configuration and start the dashboard at http://localhost:3200. A dashboard‑only mode requires no code editor—just npx patchwork-os@beta init then patchwork start.

Key features of Patchwork OS

  • 177 built-in tools (LSP, git, terminal, debugger, files) plus hot-reloadable plugins
  • YAML recipes triggered by cron, file save, git commit, test run, or webhook
  • Delegation policy with three risk tiers and four‑source precedence
  • Durable trace memory – every approval and recipe run is saved as JSONL
  • OAuth with PKCE S256 and dynamic client registration for a private personal API
  • Works natively on Windows, macOS, and Linux with no WSL required

Use cases of Patchwork OS

  • Morning Brief: AI digest of Gmail, calendar, and tasks delivered to an inbox
  • Automated CI triage – analyse test failures and write a report before you wake up
  • Draft email replies in your voice pending human approval
  • Run background agent workflows that ask before modifying anything sensitive
  • Connect multiple AI clients (Claude Code, Gemini CLI, Grok Build) to the same runtime

Frequently asked questions

1

What are the runtime requirements for Patchwork OS?

Node.js 22+ is required. On macOS/Linux `tmux` is auto‑detected but the runtime falls back to background mode if absent. Windows is natively supported (no WSL needed) for the bridge and dashboard.

2

Where are my data, logs, and configuration stored?

All configuration, inbox files, and trace logs live in `~/.patchwork/`. The trace memory is durable JSONL that you can bundle and back up with `patchwork traces export`.

3

Can I use Patchwork OS without Claude Code or any code editor?

Yes. The dashboard runs standalone as a Next.js app over HTTP. You can schedule YAML recipes, manage approvals, and read the AI inbox entirely from the browser—no editor required.

4

Does Patchwork OS phone home or send telemetry?

No. Nothing phones home unless you explicitly opt in to anonymous analytics. The runtime and all data stay on your machine.

5

How do I connect Gemini CLI or Grok Build to Patchwork OS?

Gemini CLI connects via Streamable HTTP with a Bearer token (use `patchwork-os print-token`). Grok Build connects via a stdio shim (`claude-ide-bridge shim --workspace /path`) or Streamable HTTP. See the README for exact configuration steps.

Comments

More Productivity MCP servers