1. Conduid
  2. Developer Tools
  3. Claude Colony
MCP server · Developer Tools

Claude Colony

Multi-agent Claude Code you can actually control. Spawn specialist teams in tmux, watch them work side-by-side, steer when needed. Built on native Claude Code agents. No black boxes.

66Good

Scored 22 hours ago · breakdown

About Claude Colony

Claude Colony is an MCP server published by MakingJamie in the Developer Tools category: multi-agent Claude Code you can actually control. Spawn specialist teams in tmux, watch them work side-by-side, steer when needed. Built on native Claude Code agents. No black boxes. It has been installed 0 times through Conduid.

The repository has 10 stars and 1 forks, with the last commit 8 months ago. Six months or more without a commit doesn't mean the server is broken, but check the open issues (0) before depending on it in production.

Install

Install
npx claude-colony

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

Ask AI about Claude Colony

Powered by Claude · Grounded in docs

I know everything about Claude Colony. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

  • ·README presentNot checked yet.
  • ·License declaredNot checked yet.
  • ·Tests presentNot checked yet.
  • ·Dependencies pinnedNot checked yet.
  • ·No dynamic code executionNot checked yet.
  • !Scoped permissionsDoesn't declare a permission scope. Assume it can do anything its process can.

Releases

v0.1.1v0.1.1

README

Claude Colony

npm version

A tmux-based multi-agent orchestration system where multiple Claude Code instances collaborate on software development tasks.

Overview

Claude Colony enables a team of AI agents to work together on complex tasks. It builds on Claude Code's native agent system (.claude/agents/) — create your agents there, then orchestrate them as a team.

Each agent has a unique persona, role, and area of focus. They communicate via a file-based message broker and are visualized in a tmux session.

┌─────────────────────┬─────────────────────────────┐
│                     │      Specialist Agent 1     │
│                     ├─────────────────────────────┤
│   Manager Agent     │      Specialist Agent 2     │
│      (40%)          ├─────────────────────────────┤
│                     │      Specialist Agent 3     │
│                     ├─────────────────────────────┤
│                     │   Shared Context (10%)      │
└─────────────────────┴─────────────────────────────┘

Claude Colony in action

Why Claude Colony?

True Multi-Agent Full Claude Code instances, not sandboxed sub-tasks. Each agent can spawn its own agents, use every tool, and work in parallel—just like a real dev team
Native Integration Built on Claude Code's agent system — your .claude/agents/ work everywhere
Visual Collaboration Watch your AI team work in real-time via tmux
Simple Communication File-based @mention protocol, no complex setup
Ready-Made Teams Start with fullstack, frontend, API, or mobile presets instantly

Note: Because each agent is a full Claude Code instance with unlimited capability, token usage can add up quickly. We recommend using Claude Colony with a Max plan for the best experience.

Prerequisites

  • Node.js 18+
  • tmuxbrew install tmux (macOS) or apt install tmux (Linux)
  • Claude Code — installed and authenticated
# Verify Claude Code is working
claude --version

Installation

# Install globally from npm
npm install -g claude-colony

# Or run directly with npx
npx claude-colony --help

From Source

git clone https://github.com/jamiemurphy/claude-colony.git
cd claude-colony
npm install && npm run build && npm link

Quick Start

# Initialize colony in your project
colony init

# Start a colony with a preset
colony start --preset fullstack

# Check status
colony status

# Stop the colony
colony stop

Tip: For best results, create custom agents in .claude/agents/ tailored to your project, then build a preset that uses them with colony create-team.

CLI Commands

Command Description
colony init Initialize .colony/ directory in current project
colony start Start a colony session
colony stop Stop the running colony
colony status Show colony status
colony presets List available presets
colony presets <name> Show preset details
colony agents List available agents
colony agents <name> Show agent details
colony create <name> Create a new agent
colony build Initialize and configure colony (init + setup)

Start Command Options

colony start [options]

Options:
  -p, --preset <name>   Use a preset configuration
  -a, --agents <count>  Number of worker panes (default: 4)
  --no-attach           Don't attach to tmux session

Create Command Options

colony create <name> [options]

Options:
  -d, --description <desc>  Agent description
  -m, --model <model>       Model (sonnet, opus, haiku)
  -t, --tools <tools>       Comma-separated tools
  -g, --global              Create in ~/.claude/agents/
  -f, --force               Overwrite existing

Presets

Built-in presets for common team configurations:

Preset Description Agents
fullstack Full-stack web development UX, Frontend, Backend, DevOps
frontend-team UI-focused development UX, UI Dev, Logic Dev
api-team Backend API development API Dev, Data Dev, Docs
mobile-app Mobile app development UX, Mobile Dev, API Dev, DevOps
docs-team Documentation team API Docs, Guides, Review
# List all presets
colony presets

# View preset details
colony presets fullstack

# Start with a preset
colony start --preset fullstack

Workflow Examples

Feature Development

colony start --preset fullstack

# Agents coordinate via @mentions:
# @ux:       "I'll design the login flow wireframes"
# @frontend: "Building the form based on UX specs"
# @backend:  "Creating /auth endpoints"
# @devops:   "Setting up CI for the new routes"

API Sprint

colony start --preset api-team

# Parallel development:
# @api-dev builds endpoints while @data-dev designs schemas
# @docs documents as features complete

Debug Session

colony start -p fullstack

# In the shared pane, give context:
# @all: "Users report 500 errors on checkout. @backend investigate
#        the /orders endpoint, @frontend check error handling"

Agents

Claude Colony uses Claude Code's native agent system. The recommended workflow is:

  1. Create agents in Claude Code using .claude/agents/*.md files
  2. Use those agents in your colony presets

Creating Agents (Recommended)

Create agents directly in your project's .claude/agents/ directory. These are standard Claude Code agents that work both standalone and in colonies:

<!-- .claude/agents/api-designer.md -->
---
name: api-designer
description: REST API design specialist
model: sonnet
---

# API Designer

You are an API design specialist focused on RESTful best practices...

Agent Locations

Location Scope Description
.claude/agents/*.md Project Recommended. Committed to git, works with Claude Code
~/.claude/agents/*.md User Personal agents, available across all projects
Built-in Colony Fallback defaults (frontend-dev, backend-dev, etc.)

Using Your Agents

# See all available agents (from .claude/agents/ and built-ins)
colony agents

# View agent details
colony agents api-designer

# Create a preset that uses your agents
colony create-team

Quick Agent Creation

If you need a quick agent without writing markdown:

colony create my-agent -d "My custom agent" -m sonnet

This creates a .claude/agents/my-agent.md file you can then customize.

Project Structure

After running colony init:

.colony/
├── config.yaml    # Colony configuration
├── state/         # Session state
├── messages/      # Agent communication
├── agents/        # Agent prompts
├── shared/        # Shared data channels
└── reviews/       # Code reviews

Architecture

Core Components

  • MessageBroker - File-based IPC with @mention routing
  • SessionManager - Session state persistence
  • TmuxManager - Tmux session and pane management
  • CompletionDetector - Monitors worker completion
  • ColonyMonitor - Real-time status monitoring
  • SharedScratchpad - Typed data channels for agent collaboration
  • OwnershipRegistry - File ownership tracking
  • ReviewManager - Code review workflow
  • DependencyGraph - Task dependency tracking
  • FileLock - Concurrent access safety

Communication Protocol

Agents communicate using @mentions:

@frontend: Build the login form with validation
@backend: Create the /auth/login endpoint
@all: Team standup in 5 minutes
@manager: I'm blocked on the API schema

Agent Status

  • 🟢 Ready — awaiting work
  • 🟡 Working — actively processing
  • 🔴 Blocked — needs input/help
  • ✅ Done — declared completion

Viewing the Colony

Attach to the tmux session:

tmux attach -t colony

Navigation:

  • Ctrl+B then Arrow Keys - Switch panes
  • Ctrl+B then D - Detach (colony keeps running)

How It Compares

Claude Colony Single Claude Code Generic Orchestrators
Agent architecture Full instances Single instance Task-based sub-agents
Agents spawn sub-agents Yes Yes Limited/No
Parallel execution Yes No Yes
Native .claude/agents/ Yes Yes No
Visual monitoring tmux real-time Terminal Varies
Agent communication @mention protocol N/A Custom
Setup colony init N/A Often complex

Development

# Install dependencies
npm install

# Build
npm run build

# Run tests
npm test

# Run tests once
npm run test:run

# Type check
npm run typecheck

# Lint
npm run lint

# Full check
npm run check

License

MIT

README mirrored from the source repository 22 hours ago. The original is authoritative.

Questions

About Claude Colony

How do I install Claude Colony?

Run npx claude-colony, then add the server to your MCP client's configuration. Conduid has recorded 0 installs, so the command is known to work with current clients.

Is Claude Colony safe to use with an AI agent?

Its trust score is 66 out of 100 (good). It passes 0 of 1 static security checks; the failures are listed above. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Claude Colony still maintained?

The last commit was 8 months ago, with 0 open issues. That's long enough that you should check whether the maintainer is responding to issues before depending on it.