RE
OfficialFinance

Revit MCP Server AI Connector for Element Selection

Connect AI Desktop apps to Revit with the MCP server to inspect and select elements using 36 predefined tools.

View on GitHub

Overview

The Revit MCP Server (AI Connector for Revit) bridges AI desktop applications (Claude, Cursor, Copilot, etc.) with Autodesk Revit without requiring the AI to emit Revit API code. Instead, the connector exposes a fixed set of 36 inspection-and-selection micro-tools that the AI can invoke through the Model Context Protocol (MCP). Each tool runs inside Revit, returns model feedback, and optionally performs read-only selection actions — reducing risk and providing reliable, reproducible operations.

This approach is useful when you want conversational AI to inspect, summarize, or select elements in a Revit model (for review, QA, or downstream automation) while avoiding the brittleness of generated Revit code. The connector is compatible with Revit 2022–2026, requires the Nonicatab plugin, and works with MCP-capable desktop AI apps.

Features

  • 36 predefined, read-only tools for model inspection and element selection
  • Tools to read parameters, family sizes, sheets, views, viewports, and element geometry metadata
  • Feedback-driven interaction: tools return structured model data the AI uses to adapt follow-up queries
  • Read & Select only (Phase 1): no direct model edits, only model reading and element selection
  • Integrates into Nonicatab toolbar in Revit for easy activation
  • Works with Claude Desktop, Cursor, Copilot in VSCode, and other MCP-compatible clients

Installation / Configuration

Prerequisites:

  • Autodesk Revit 2022, 2023, 2024, 2025, or 2026
  • Nonicatab Revit plugin installed
  • An MCP-capable AI desktop app (Claude Desktop, Cursor, Copilot)

Steps:

  1. Download and install the AI Connector for Revit from the Autodesk App Store or the GitHub repository.
  2. Install/enable Nonicatab in Revit and ensure the Nonicatab toolbar is visible.
  3. Open your MCP-capable AI desktop app (Claude, Cursor, Copilot).
  4. Launch Revit and open the Nonicatab toolbar; select the ready-to-use AI Connector toolbar or tool.
  5. Start the AI Connector in Revit — the MCP server will become available to your AI app.

Example MCP tool invocation (pseudo-JSON request sent by the AI client):

{
  "mcp_action": "run_tool",
  "tool_name": "get_parameters_from_elementid",
  "arguments": {
    "list_elementId": [123456, 123457],
    "name": "all"
  }
}

Example response (structured result containing parameters and selection info):

{
  "status": "ok",
  "result": [
    {
      "elementId": 123456,
      "parameters": {
        "Family": "Door - Single Flush",
        "Width": "900 mm",
        "Material": "Steel"
      }
    }
  ],
  "selection_changed": true,
  "selected_ids": [123456]
}

Security note: Starting an AI desktop app with MCP access grants that app the ability to inspect your open Revit model. Review privacy and access policies for any AI provider used.

Available Tools (selected)

There are 36 micro-tools in total. Below are representative tools — use your MCP client’s tool list or the AI’s Search & Tools view to see the full set.

Tool namePurpose
get_parameters_from_elementid(list_elementId, name)Return parameters for specified element IDs
extract_size_in_MB_of_families(familyName)Report family file sizes (MB) for given family name
get_viewports_placed_on_sheets(sheetNumber)List viewports and their view names on a sheet
list_elements_by_category(categoryName)Return IDs & basic props for elements of a category
get_elements_by_parameter(cat, paramName, value)Find elements where a parameter matches a value
get_family_types_for_family(familyName)List types available for a family and type metadata

Use Cases

  • QA: “List all doors with a Width less than 800 mm and select them so I can review in the model.”
    Workflow: AI calls get_parameters_from_elementid or get_elements_by_parameter, returns list and selects IDs in Revit for visual inspection.

  • Performance review: “Which families are larger than 1 MB? Provide family names and sizes.”
    Workflow: AI invokes extract_size_in_MB_of_families to produce a ranked list of heavy families for optimization.

  • Drawing and sheet checks: “Show all views placed on sheet A101 and select the viewports.”
    Workflow: AI calls get_viewports_placed_on_sheets, returns view names and viewport IDs; selection highlights viewports in Revit.

  • Cross-check parameters: “Are there any walls with FireRating missing? Select them.”
    Workflow: AI runs get_elements_by_parameter with paramName=FireRating, value=null and selects results for manual correction.

Developer tips

  • Start by running a small query (single element ID) to verify the connector and ACLs are working before broad queries.
  • Use the AI client’s tool explorer to discover available arguments and example return formats.
  • Remember the connector is read-only in Phase 1 — use it for inspection, reporting, and selection workflows; editing automation is not supported yet.

Resources

  • GitHub: https://github.com/NonicaTeam/AI-Connector-for-Revit — source, issues, and contribution notes
  • Autodesk App Store listing for installer and release notes
  • Nonicatab plugin (required) — enable the AI Connector toolbar in Revit

For questions about tool behavior or to request additional micro-tools, open an issue on the GitHub repository.

Tags:finance