v0.1.0

idfkit-mcp v0.1.0

First public release of the MCP server for EnergyPlus model authoring, validation, and simulation via idfkit.

Highlights

25+ MCP Tools

• Model I/O: load_model, new_model, save_model, convert_osm_to_idf, get_model_summary
• Editing: add_object, update_object, duplicate_object, batch_add_objects, remove_object, rename_object
• Schema: list_object_types, describe_object_type, search_schema, get_available_references
• Validation: validate_model, check_references
• Simulation: run_simulation, get_simulation_summary, get_error_report, list_output_variables, query_timeseries, export_timeseries
• Weather: search_weather_stations, download_weather_file
• Documentation: lookup_documentation, search_docs, get_doc_section
• Session: clear_session

Structured Output

All tools return Pydantic-validated response models with full JSON schemas, enabling reliable structured output for AI clients.

MCP Tool Annotations

Every tool includes readOnlyHint, destructiveHint, idempotentHint, and openWorldHint annotations for safe autonomous operation.

Session Persistence

Loaded models, simulation results, and weather files survive server restarts — critical for clients like Codex that restart the MCP server between turns.

Documentation Integration

• doc_url fields on describe_object_type, search_schema, and validate_model responses link directly to docs.idfkit.com
• Full-text documentation search via search_docs and get_doc_section

Platform Support

Tested with Claude Desktop, Cursor, VS Code Copilot, Windsurf, Cline, OpenAI Codex, and more. Supports stdio, SSE, and streamable-http transports. Docker image available.

Requirements

• Python 3.10+
• idfkit >= 0.6.0

Install

# pip / uvx
pip install idfkit-mcp

# Claude Desktop (claude_desktop_config.json)
{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["idfkit-mcp"]
    }
  }
}