MCP for AI agents

MCP for AI agents

Use the PyOrchestrator MCP server with Cursor and other AI agents to list scripts, run jobs, read logs, and manage schedules.

PyOrchestrator provides an MCP server (mcp/) to manage the platform of AI agents (Cursor, Claude, custom bots).

Transport

Mode When to use
stdio Local development, Cursor IDE
streamable-http Docker Compose, port 8010

Tools (24)

Category Tools
Auth pyorch_login, pyorch_whoami
Scripts list_scripts, get_script, create_script, update_script_file, enable_script, disable_script, delete_script
Runs run_script, stop_script, get_run, get_run_logs, list_script_runs
Automation list_groups, list_schedules, create_schedule, list_webhooks, create_webhook
Secrets set_script_secret, list_script_secrets
Platform dashboard_stats, system_info, list_notifications

Resource: pyorch://platform/overview

Cursor (stdio)

{
  "mcpServers": {
    "pyorchestrator": {
      "command": "python3",
      "args": ["-m", "pyorchestrator_mcp"],
      "cwd": "/path/to/PyOrchestrator/mcp",
      "env": {
        "PYORCH_API_URL": "http://localhost:8000",
        "PYORCH_EMAIL": "admin@pyorchestrator.local",
        "PYORCH_PASSWORD": "admin"
      }
    }
  }
}

Docker (HTTP)

docker compose up -d mcp

URL: http://localhost:8010/mcp

{
  "mcpServers": {
    "pyorchestrator": {
      "url": "http://localhost:8010/mcp"
    }
  }
}

Environment variables

Variable Description
PYORCH_API_URL Backend API
PYORCH_TOKEN JWT (optional)
PYORCH_EMAIL / PYORCH_PASSWORD Autologin
MCP_TRANSPORT stdio | streamable-http
MCP_PORT HTTP port (default 8010)

Example workflow agent

  1. system_info - health check
  2. list_scripts - find the script
  3. update_script_file - change main.py
  4. run_script → get_run_logs - execute and read the output

More details: mcp/README.md