npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

coworker-mcp

v0.1.0-alpha.7

Published

Turn Cowork into an autonomous PM for Claude Code

Readme


Coworker is an open-source local MCP server that lets Anthropic's Cowork (or any MCP client) delegate coding tasks to Claude Code — with async parallel execution, auto-verification, session-resuming iteration, and token-efficient summaries that keep the client's context window small.

93-98% fewer tokens in the orchestrator's context. Zero copy-paste. Parallel task execution. Your AI PM thinks while your AI engineer codes.

The Problem

You use Cowork for thinking and Claude Code for coding. But every task requires you to:

  1. Copy context from Cowork
  2. Paste it into Claude Code
  3. Wait for results
  4. Copy the output
  5. Paste it back to Cowork
  6. Repeat 10-15 times per feature

You're a human API between two AI tools. Coworker eliminates the middleman.

How It Works

You ── natural language ──► Cowork (PM)
                               │
                               │ MCP tools (submit, iterate, verify)
                               ▼
                          Coworker (bridge)
                               │
                               │ spawns subprocess
                               ▼
                          Claude Code (engineer)
                               │
                               │ reads/writes
                               ▼
                          Your project files
  1. You describe what you want to Cowork
  2. Cowork calls submit_task through the MCP bridge
  3. Coworker spawns claude -p as a subprocess on your machine
  4. Claude Code does the work using your local files and tools
  5. Coworker captures the output, runs verification, builds a summary
  6. Only the summary returns to Cowork — full output stays on disk
  7. For iteration, Cowork calls iterate_task — Claude Code resumes its session and remembers everything

Real Numbers

Measured from a 5-task development session with iterations:

| Metric | Without Coworker | With Coworker | |--------|-----------------|---------------| | Tokens in PM context | ~15,000-55,000 | ~1,100 | | User actions per task | 4 (copy, switch, paste, switch) | 0 | | Parallel tasks | Impossible | Yes | | Claude Code cost | $0.62 | $0.62 (identical) | | Context runway | ~5-8 tasks before degradation | 50+ tasks |

Coworker doesn't save money on the coding side — it saves your time and your PM's context space.

Quick Start

Cowork plugin (recommended)

One-click Cowork install. Bundles the MCP connector + 2 skills. See plugin/README.md.

One-command setup

npx coworker-mcp setup

This checks dependencies, downloads cloudflared if needed, initializes the project, and starts the server. The connector URL is automatically copied to your clipboard.

Then connect to Cowork

  1. Open Claude Desktop → Settings → Connectors
  2. Add custom connector → Paste (URL is on your clipboard)
  3. Save → Toggle on in your conversation
  4. Say: "What tools do you have?" — you should see 6 Coworker tools

For a permanent URL (recommended)

npx coworker-mcp setup --stable

Sets up a named Cloudflare tunnel. URL never changes. Paste once, done forever.

Requirements

  • Node.js 20+
  • Claude Code CLI (authenticated)
  • cloudflared (auto-downloaded if missing)

MCP Tools

submit_task

Submit a coding task to Claude Code. Returns immediately — Claude Code runs asynchronously in the background.

wait_for_task

Wait for a running task to complete. Returns the summary once done. Enables the submit → do other things → check result pattern.

get_result

Fetch a task's result at three detail levels: oneline, paragraph, or full (returns file path only — output never enters the client's context).

iterate_task

Continue a task with feedback. Resumes Claude Code's prior session via --resume — it remembers all previous work. No context re-sent.

list_tasks

List recent tasks with compact summaries. Filter by status, time range, or search prompt text.

get_project_state

Get the current project state: what's been built, project context, and decisions. Start any new conversation with this to catch up instantly.

Auto-Verification

Coworker can run your tests and build commands after every task. If they fail, it automatically feeds the error back to Claude Code and retries.

# .coworker/config.yaml
verification:
  enabled: true
  commands:
    - npm test
    - npm run lint
  max_retries: 2

You get results that already pass your checks.

Project State

Coworker maintains three files automatically — you never edit these:

| File | Purpose | Updated | |------|---------|---------| | STATUS.md | What's been built, what failed | After every task | | CONTEXT.md | Project description, tech stack | By you or Cowork | | DECISIONS.md | Architecture decisions log | During conversations |

Start a new Cowork conversation → ask to read the project state → fully caught up in one message.

CLI Reference

coworker setup [--stable]    One-command onboarding (--stable for permanent URL)
coworker init [dir]          Initialize .coworker/ in a project
coworker start [--port N]    Start the MCP server and tunnel
coworker url                 Print connector URL (auto-copies to clipboard)
coworker tunnel-setup        Set up a permanent tunnel URL
coworker history [--limit N] List recent tasks
coworker show <task_id>      Show task details and summary
coworker doctor              Run health checks (9 checks)

Configuration

All fields optional. Sensible defaults for everything.

# .coworker/config.yaml
version: 1

claude:
  binary_path: claude
  default_timeout_seconds: 600
  default_max_turns: 20
  working_directory: .

server:
  port: 17429
  tunnel_mode: quick            # quick | named | none

summary:
  mode: heuristic               # heuristic (free) | llm (~150 tokens/task)

verification:
  enabled: false
  commands: []
  max_retries: 2

limits:
  max_concurrent_tasks: 5
  max_task_age_hours: 24

FAQ

Security

Coworker runs Claude Code with --dangerously-skip-permissions. Recommendations:

  • Only run in project directories you trust
  • Use allowed_tools in submit_task to restrict capabilities
  • Review .coworker/results/ for full audit trail
  • Consider running in a container for isolation

Built With

License

MIT