mcp s-for-revit

The mcp-server-for-revit provides a TypeScript MCP server that exposes a set of Revit-related tools to AI assistants via the Model Context Protocol. It acts as a bridge between an MCP client (like Claude or other MCP-compatible agents) and the Revit environment by translating tool calls into WebSocket messages that the Revit plugin and command set can execute inside Revit. Once the server is running, AI clients can request operations such as reading the current view, creating or modifying elements, or querying model data through the published toolset. The server itself does not perform Revit operations directly; instead, it forwards requests to the Revit plugin, which dispatches to the Command Set that interfaces with the Revit API and returns results back through the same channel.

To use it, install/run the server package via npx as documented, then configure your MCP client to connect to the server under a named entry (for example, mcp-server-for-revit). The available tools include tasks like get_current_view_info, create_point_based_element, delete_element, tag_all_rooms, export_room_data, and many others listed in the Supported Tools section. Your AI assistant can call these tools by name with appropriate parameters, and the results (such as IDs, coordinates, or JSON data) will flow back to the AI client for interpretation and response generation.

Prerequisites:

• Node.js 18+ installed on your machine
• Access to a Revit environment (2020–2026) for runtime testing

Installation steps:

1. Ensure Node.js is installed (version 18 or newer).
2. Open a terminal and install the MCP server package using npx (no global install required): npm install -g npm@latest npx -y mcp-server-for-revit
3. (Optional) If you want to pin a specific version, replace the package name with a version specifier, e.g. mcp-server-for-revit@1.0.0.
4. Prepare your Revit side following the repository's Quick Start: download the appropriate release ZIP for your Revit version, extract, and place into the Revit Addins folder as described in the README.
5. Ensure the Revit plugin (revit_mcp_plugin) and the Command Set (RevitMCPCommandSet) are present in the Addins directory, as outlined in the manual installation steps.
6. Configure your MCP client to connect to the server (see the client configuration example in the README).

• The MCP server is designed to work with Claude, Claude-like assistants, and other MCP-compatible clients. The example configuration shown uses the server name mcp-server-for-revit; adjust the server entry name to match your setup.
• Ensure the Revit plugin loads automatically when Revit starts; if you install manually, follow the step-by-step plugin setup in the README.
• The available tools cover a broad range of Revit operations, from querying view information to creating and editing elements. When constructing tool calls, provide the required parameters per tool (types and shapes vary by operation).
• If you encounter connection issues, verify that the WebSocket channel between the MCP server and the Revit plugin is open and that the correct tool names are used by the AI client.
• For testing within Revit, you may use the included testing setup in the repository or follow the integration guidance in the README to run tests against a live Revit instance.