pddl

This MCP server provides a Natural Language to PDDL planning workflow. It accepts natural language inputs, generates corresponding PDDL planning problems, executes plans using the integrated Fast Downward planner, and supports batch processing to produce detailed execution reports. The server exposes capabilities for NLP-driven task formulation, automatic problem generation, plan execution, and batch task orchestration, with type-safe configuration validation via Pydantic. To use it, start the server and connect via MCP-compatible clients (e.g., Claude Desktop or Trae AI IDE). The server will generate PDDL domains and problems, run planners, and output plans, explanations, and batch results to the configured output directory.

Prerequisites:

• Python 3.8+ installed
• access to a compatible Fast Downward installation
• network access to install Python dependencies

Step-by-step:

1. Install Python dependencies

   pip install -r requirements.txt
   

2. Prepare Fast Downward
   • Clone and build Fast Downward:

     git clone https://github.com/aibasel/downward.git
     cd downward
     ./build.py
     

3. Configure environment variables
   • Copy example env and edit paths:

     cp .env.example .env
     # Edit .env
     FAST_DOWNWARD_PATH=/path/to/fast-downward.py
     PDDL_DOMAIN_PATH=./templates/domain.pddl
     

4. Run the MCP server

   python server.py
   

5. Validate installation
   • Access logs and ensure the server reports initialization and FastMCP startup messages.

Notes:

• Ensure you have network access for dependencies and that the output directory exists and is writable.
• The PDDL domain template path should point to a valid domain file used by the planner.

Tips and common issues:

• If the server fails to start, verify Python version (3.8+), check that requirements are installed, and confirm the .env configuration is loaded.
• Ensure FAST_DOWNWARD_PATH points to a valid fast-downward.py or launcher script within your Downward checkout.
• The PDDL_DOMAIN_PATH should reference an existing domain template used to generate problems; mismatches can cause planning failures.
• When running batch tasks, monitor the output directory for batch reports and ensure there is sufficient disk space for generated plans and logs.
• For Trae IDE or Claude Desktop integration, ensure the MCP server config matches the expected fields (name, command, args, cwd, env).
• If you modify templates or task configurations, re-run the system config check to validate types and paths.