Karajan Code

v1.0.0Coding Agentsstable

Local multi-agent coding orchestrator with 16 pipeline roles, TDD enforcement, SonarQube integration, and automated code review. Supports Claude, Codex, Gemini, OpenCode and Aider. MCP server included. No API costs — runs on your existing subscriptio

ai-agentai-orchestrationorchestrator
Share:
11
Stars
0
Downloads
0
Weekly
0/5

What is Karajan Code?

Karajan Code is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to local multi-agent coding orchestrator with 16 pipeline roles, tdd enforcement, sonarqube integration, and automated code review. supports claude, codex, gemini, opencode and aider. mcp server included...

Local multi-agent coding orchestrator with 16 pipeline roles, TDD enforcement, SonarQube integration, and automated code review. Supports Claude, Codex, Gemini, OpenCode and Aider. MCP server included. No API costs — runs on your existing subscriptio

This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • Local multi-agent coding orchestrator with 16 pipeline roles

Use Cases

Orchestrate multi-agent coding with TDD enforcement.
Automate code review and integrate SonarQube analysis.
manufosela

Maintainer

LicenseAGPL-3.0
Languagejavascript
Versionv1.0.0
UpdatedMay 18, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx karajan-code

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Karajan Code

Karajan Code is a local multi-agent coding orchestrator that coordinates up to 18 AI-backed pipeline roles — from triage and planning through coding, SonarQube analysis, security audit, and test verification — across Claude, Codex, Gemini, Aider, and OpenCode, with zero API costs beyond your existing subscriptions.

Prerequisites

  • Node.js >= 22.22.1 (active LTS) installed
  • Git repository with a configured remote
  • At least one AI agent CLI installed: Claude Code (npm install -g @anthropic-ai/claude-code), Codex, Gemini CLI, Aider, or OpenCode
  • Docker (optional, required for Ollama embeddings and SonarQube integration)
  • Minimum 4 GB RAM and ~250 MB disk space; 8–12 GB RAM recommended for full pipeline
1

Install the kj Binary

Install Karajan Code globally via npm. Alternatively use Homebrew on macOS, download a standalone binary (no Node.js required), or run the one-line shell installer.

# npm (recommended)
npm install -g karajan-code

# Homebrew (macOS)
brew install manufosela/tap/karajan-code

# One-liner installer
curl -fsSL https://raw.githubusercontent.com/manufosela/karajan-code/main/scripts/install-kj.sh | sh
2

Bootstrap Karajan in Your Project

Run kj init inside your Git repository. The wizard auto-detects installed AI agents, frameworks, and optional services (Docker, SonarQube), and writes a .karajan/kj.config.yml project config. It also asks once whether to enable anonymous telemetry (off by default).

cd /path/to/your-project
kj init
3

Verify the Environment

Run kj doctor to confirm all detected agents and optional dependencies are correctly wired. This surfaces missing CLI tools or mis-configured paths before you run a real pipeline.

kj doctor
4

Run Your First Pipeline

Pass a plain-English task description to kj run. Karajan orchestrates the full pipeline: intent classification, planning, coding (TDD-first), SonarQube gate, review, and test verification. For quick coding without review use kj code.

# Full pipeline
kj run "Add input validation to the user registration endpoint"

# Coder only (skips review loop)
kj code "Fix the off-by-one error in the pagination helper"
5

Configure the MCP Server

Karajan auto-registers its MCP server after installation. If manual setup is needed, add the following block to your Claude config file (~/.claude.json). This exposes 27 tools including kj_run, kj_code, kj_review, kj_plan, kj_audit, and kj_rag_query.

{
  "mcpServers": {
    "karajan-mcp": {
      "command": "karajan-mcp"
    }
  }
}
6

Tune Agent Routing via kj.config.yml

Create or edit .karajan/kj.config.yml in your project to assign different AI agents to specific pipeline roles and enable optional stages like the architect, refactorer, or commiter.

agents:
  coder: claude
  reviewer: codex

pipeline:
  roles:
    planner: true
    researcher: true
    architect: false
    refactorer: false
    commiter: false

rag:
  preload:
    enabled: true
    topK: 5
  embedder:
    provider: ollama

telemetry: false

Karajan Code Examples

Client configuration

Claude Desktop or Claude Code config to connect to the Karajan MCP server, which exposes 27 pipeline tools.

{
  "mcpServers": {
    "karajan-mcp": {
      "command": "karajan-mcp"
    }
  }
}

Prompts to try

Example tasks to send through Claude once Karajan MCP is connected, or directly via the kj CLI.

- "Run the full Karajan pipeline to add rate limiting to the API."
- "Review the current git diff for security issues using the Karajan audit role."
- "Generate a structured implementation plan for adding OAuth2 support."
- "Use kj_rag_query to find all patterns related to error handling in this codebase."
- "Resume the interrupted pipeline with ID listed in kj plan list."

Troubleshooting Karajan Code

kj doctor reports 'No agent found' even though Claude Code is installed

Ensure the agent CLI is on your PATH. Run 'which claude' to confirm. If it is not found, reinstall with 'npm install -g @anthropic-ai/claude-code' and open a new terminal session before running kj doctor again.

SonarQube gate fails immediately with connection refused

SonarQube runs in Docker. Start it with 'docker compose up -d sonarqube' from the Karajan home directory, wait for it to become healthy (can take 60–90 seconds), then retry. Disable the sonar role in kj.config.yml if you do not want SonarQube.

Pipeline is interrupted mid-way due to agent quota limits

Karajan supports hibernation. Run 'kj plan list' to find the interrupted pipeline ID, then resume with 'kj resume <id>'. The pipeline restarts from the last completed stage.

Frequently Asked Questions about Karajan Code

What is Karajan Code?

Karajan Code is a Model Context Protocol (MCP) server that local multi-agent coding orchestrator with 16 pipeline roles, tdd enforcement, sonarqube integration, and automated code review. supports claude, codex, gemini, opencode and aider. mcp server included. no api costs — runs on your existing subscriptio It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Karajan Code?

Follow the installation instructions on the Karajan Code GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Karajan Code?

Karajan Code works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Karajan Code free to use?

Yes, Karajan Code is open source and available under the AGPL-3.0 license. You can use it freely in both personal and commercial projects.

Browse More Coding Agents MCP Servers

Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "karajan-code": { "command": "npx", "args": ["-y", "karajan-code"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Karajan Code?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides