This MCP server exposes a JSON-RPC style API that lets external AI agents, CLIs, and other tools interact with Autodesk Revit through a dedicated Revit add-in. It acts as a local IPC/HTTP entry point (depending on its runtime implementation) and translates incoming requests into Revit API calls executed by the add-in. Typical capabilities include querying element lists by category, retrieving parameters, editing parameter values, moving or rotating elements, and logging operation histories. The server is designed to be used from automation pipelines or AI agents that require safe, auditable access to Revit data and operations, with commands defined in the OpenAPI/JSON-RPC definitions referenced by the project.

To use it, run the server as described in the installation guide, ensure the host has a valid Autodesk Revit license, and connect via the configured RPC endpoint (default localhost:5000). Your client can then send requests like “getElements”, “updateParameter”, or “moveElement” following the API schema. Because the tool supports safety checks and dry-run capabilities, you can validate changes before applying them to a live Revit model. Integration with your existing workflow (e.g., AI agent prompts, CLI tools, or other automation scripts) is supported through standard JSON-RPC transports and local IPC/HTTP as implemented in the server.

Prerequisites:

• Autodesk Revit license and a compatible Windows environment
• Python 3.8+ installed on the host (or a suitable runtime if the project uses a packaged distribution)
• Access to the repository containing revittoolkit (this MCP server)
• Internet access for initial package installation (if dependencies are external)

Installation steps:

1. Clone the repository containing the RevIt MCP toolkit server: git clone https://github.com/your-org/revit-mcp-toolkit.git cd revit-mcp-toolkit

2. Set up a Python virtual environment (recommended): python -m venv venv

   Windows

   venv\Scripts\activate.bat

   macOS/Linux

   source venv/bin/activate

3. Install dependencies (adjust if a requirements file exists): pip install -r requirements.txt

   or if using poetry

   poetry install

4. Configure environment (as needed):

   • Ensure REVIT_LICENSE is available on the host
   • Optionally adjust RPC_BIND_ADDRESS and RPC_BIND_PORT in env variables

5. Run the MCP server: python -m revittoolkit.server

   For production, consider a supervisor/service wrapper and logging

6. Verify the server is up by hitting the RPC endpoint (default http://127.0.0.1:5000 or as configured) and performing a basic request (e.g., a dry-run query if supported).

7. Optional: containerized run (if provided by the project): docker run -i your-docker-registry/revit-mcp-toolkit:latest

Notes:

• If your environment uses a different runtime (e.g., a compiled .NET-based runner), adapt the command accordingly to invoke the actual server binary/executable.
• Ensure network/firewall settings allow IPC/HTTP traffic on the chosen port.

Tips and common considerations:

• Always use a dry-run or confirm-flow for critical operations to minimize unintended changes in the Revit model.
• Enable access control and audit logging in the server configuration to satisfy organizational security policies.
• Keep Autodesk Revit and the MCP toolkit add-in synchronized to supported versions to avoid compatibility issues.
• If the server fails to start, check environment variables (especially RPC_BIND_ADDRESS/PORT) and the Revit license availability on the host.
• Document the available API endpoints and payload schemas for your automation teams; consider adding example requests for common tasks (e.g., retrieve all walls, modify a parameter value).
• For long-running tasks, implement task queues or asynchronous handling to avoid blocking the JSON-RPC endpoint.
• Regularly monitor logs in the configured log directory (e.g., REVOke_LOG_DIR) to detect errors early.