# omp-worker-mcp

> omp-worker-mcp is an open-source MCP server for developer tools. Delegate coding tasks from ChatGPT, Codex or Claude Code to local Oh My Pi workers. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, macOS, Windows, Linux
- **Repository:** https://github.com/divenire990/omp-worker-mcp
- **MCP Registry name:** io.github.divenire990/omp-worker-mcp
- **Categories:** [developer tools](https://ai-powered.com/categories/developer-tools), [AI models and LLM tooling](https://ai-powered.com/categories/ai-and-llm)
- **Listing updated:** 2026-11-18

## About

The idea is a split of roles: the agent you talk to acts as supervisor, who plans, writes acceptance criteria and reviews, while the Oh My Pi (OMP) CLI on your machine does the hands-on editing, shell commands and tests with the model chosen in omp setup. The server starts OMP in the background, stores each job on disk and returns a short structured report rather than raw logs.

A single task can be delegated with omp_run_compact, or up to 50 tasks can run as a dependency graph, ten at a time, with omp_run_batch_compact. Parallel write tasks must declare the paths they own, and overlapping scopes or cycles are rejected before any worker starts. The README is clear that this is a scheduling check and not an OS-level sandbox. Workers finish with an OMP_WORKER_RESULT envelope listing status, summary, changed files and verification evidence, and omp_continue sends feedback to the same session.

It runs over stdio via npx and needs Node.js 22 or newer plus OMP installed. ChatGPT on the web reaches it through OpenAI's Secure MCP Tunnel, and OMP_WORKER_ALLOWED_ROOTS (version 0.2.0 and up) restricts where jobs may run.

## Features

- **Single-task delegation:** omp_run_compact hands one task to a local OMP worker.
- **Batch dependency graph:** Up to 50 tasks run as a DAG, ten at a time.
- **Write-scope validation:** Overlapping declared write paths and cycles are rejected before work starts.
- **Structured results:** Each job ends with status, summary, changed files and verification evidence.
- **Continue same session:** omp_continue sends feedback to the worker that already has context.
- **Persistent jobs:** Records, logs and results stay on disk across server restarts.
- **Allowed roots:** OMP_WORKER_ALLOWED_ROOTS limits the directories jobs may run in.

## Good for

- Let a strong model plan while a cheaper model writes the code
- Run a multi-module refactor in parallel without overlapping edits
- Hand off a bug fix with tests and review the report
- Drive work on local files from ChatGPT on the web
- Run a long read-only investigation in the background

## Install

### Claude Code (Terminal)

```sh
claude mcp add omp-worker-mcp --env OMP_WORKER_OMP_COMMAND='<OMP_WORKER_OMP_COMMAND>' -- npx -y omp-worker-mcp
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "omp-worker-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "omp-worker-mcp"
      ],
      "env": {
        "OMP_WORKER_OMP_COMMAND": "<OMP_WORKER_OMP_COMMAND>"
      }
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "omp-worker-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "omp-worker-mcp"
      ],
      "env": {
        "OMP_WORKER_OMP_COMMAND": "<OMP_WORKER_OMP_COMMAND>"
      }
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "omp-worker-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "omp-worker-mcp"
      ],
      "env": {
        "OMP_WORKER_OMP_COMMAND": "<OMP_WORKER_OMP_COMMAND>"
      }
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "omp-worker-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "omp-worker-mcp"
      ],
      "env": {
        "OMP_WORKER_OMP_COMMAND": "<OMP_WORKER_OMP_COMMAND>"
      }
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "omp-worker-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "omp-worker-mcp"
      ],
      "env": {
        "OMP_WORKER_OMP_COMMAND": "<OMP_WORKER_OMP_COMMAND>"
      }
    }
  }
}
```

## Tools

- `omp_run_compact`: Delegate a single task to an OMP worker.
- `omp_run_batch_compact`: Run up to 50 tasks as a dependency graph.
- `omp_continue`: Send feedback to the same OMP session.

## FAQ

### What do I need installed?

Node.js 22 or newer, the Oh My Pi CLI and an MCP client such as Codex, Claude Code, Cursor or ChatGPT.

### Is it a sandbox?

No. It controls where jobs start and what they are told to touch, not what OMP is technically able to do.

### Which model do workers use?

The default model chosen in omp setup.

### How do I add it to Claude Code?

Run claude mcp add omp-worker --scope user --env OMP_WORKER_OMP_COMMAND=omp -- npx -y omp-worker-mcp.

### Is omp-worker-mcp free?

Yes. omp-worker-mcp is open source, so you can use and self-host it for free.

### How do I add omp-worker-mcp to Claude Code?

Run this command in your terminal: claude mcp add omp-worker-mcp --env OMP_WORKER_OMP_COMMAND='<OMP_WORKER_OMP_COMMAND>' -- npx -y omp-worker-mcp

### What are the best alternatives to omp-worker-mcp?

The closest alternatives to omp-worker-mcp are REA, Context7, Chrome DevTools MCP, Codebase Memory and GitHub.

## Where this comes from

- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Context7](https://ai-powered.com/mcp/context7): Context7 is a freemium MCP server for developer tools.
- [Chrome DevTools MCP](https://ai-powered.com/mcp/chrome-devtools-mcp-2): Chrome DevTools MCP is an open-source MCP server for browser automation.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [awesome-mcp-servers](https://ai-powered.com/mcp/awesome-mcp-servers-2): awesome-mcp-servers is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.

---
Source: https://ai-powered.com/mcp/omp-worker-mcp
