systemprompt-code-orchestrator

SystemPrompt Coding Agent exposes your local machine as an MCP server that MCP clients can connect to for remote, voice-enabled coding control. It runs AI-driven agents directly on your workstation with access to your real development environment and tools, enabling tasks such as code generation, refactoring, and orchestration via MCP protocol. You can monitor and drive tasks through the included inspector tool, and you can connect from any MCP-compatible client or the mobile SystemPrompt app to issue commands, track task progress, and receive notifications. The architecture also supports tunneling so you can access the server remotely when you enable Cloudflare-based tunnels.

Designed for developers, the server supports real-time resource subscriptions, task state persistence, and push notifications when long-running tasks complete. Start the daemon, optionally run it inside Docker, and use the inspector to test tasks and observe structured logs stored as MCP resources. This makes it possible to manage complex coding workflows remotely while keeping execution on your own machine for security and access to local tools.

Prerequisites:

• Node.js 18+ installed on your system
• Docker & Docker Compose (optional but recommended for docker-based workflows)
• Access to Claude Code CLI (optional but recommended; setup will guide you)

Installation steps:

1. Clone the repository

git clone https://github.com/systempromptio/systemprompt-code-orchestrator
cd systemprompt-code-orchestrator

2. Install dependencies and run initial setup

npm i
npm run setup

3. Start the MCP server (daemon + environment)

npm run start

4. Optional: run the inspector for testing and debugging

npm run inspector

5. Optional: run with Docker (if you prefer containerized deployment)

docker-compose up -d

6. Verify status and logs as needed

npm run status
npm run logs

Configuration during setup will prompt for required values and set defaults for optional ones. Ensure your PROJECT_ROOT points to the root of your code workspace where AI agents will operate, and that PORT and other env vars are set as described in the docs.

Tips and considerations:

• The MCP server implements real-time resource subscription (listChanged) for task monitoring. Ensure clients subscribe to the appropriate URIs to receive updates without polling.
• If you enable tunneling (Cloudflare), you can access the server remotely; keep your CLOUDFlARE_TOKEN secure and rotate as needed.
• The server persists tasks to disk as JSON to maintain state across daemon restarts; ensure the filesystem has adequate write permissions.
• For mobile push notifications, configure PUSH_TOKEN and verify Firebase Cloud Messaging integration.
• If you experience port conflicts, adjust PORT via environment configuration during setup or in your environment file.
• When using Docker, you can map host volumes to preserve MCP resources and logs across restarts.