pyrevit-mcp
An MCP server that exposes an open Autodesk Revit model to MCP clients, enabling natural language inspection and editing of the model via Revit API calls through pyRevit Routes.
README
pyrevit-mcp
An MCP server that exposes an open Autodesk Revit model to any MCP client, so the model can be inspected and edited in plain language instead of through the Revit UI.
Ask "find all doors narrower than 1.2m and highlight them in red" or "rename this room to ICU Recovery", and the server translates that into Revit API calls through pyRevit Routes. A small web chat UI is included for demos: split-screen with Revit, CSV import and export, undo, and screenshots of the model right in the chat.
Built for customer demos, prototyping and my own day-to-day work.
This is not an official Autodesk product and comes with no support commitment. I work at Autodesk as a Technical Specialist for Emerging Tech in Construction, and this project is not part of that role. See Disclaimer and support.
How it works
MCP client (Claude Code, Claude Desktop, ...)
│ stdio
▼
server.py (Python 3 / FastMCP) ──HTTP :48884──> pyRevit Routes (IronPython, inside Revit)
▲ │
│ shared revit_get/revit_post ▼
app.py (FastAPI :8100) ◄── web chat UI Revit API, Transactions
│
▼ Claude API
Two entry points share one bridge:
server.pyis the MCP server. Any MCP client can connect over stdio and gets 21 tools.app.pyis a FastAPI backend with a chat UI (web/index.html) driven by the Claude API. It exposes the same tools plus CSV export, and renders images, downloads, and undo in the chat.
Inside Revit, a pyRevit extension (pyrevit-mcp.extension/) registers REST routes. pyRevit
injects the active document into the handlers, which run in the Revit API context, one
transaction per write.
What it covers
21 MCP tools:
| Area | Tools |
|---|---|
| Orientation | get_status, get_model_info, get_levels, get_views, get_categories, list_parameters |
| Query | get_elements, find_elements (level + parameter filters, returns requested parameters in one call), get_element_properties |
| Write | set_parameter, set_parameters (bulk, one transaction, one undo entry) |
| Visual | override_graphics, isolate_elements, color_by_parameter (legend included), reset_view, export_image |
| Selection | get_selection, set_selection |
| Views | set_active_view |
| Model health | get_warnings (grouped, with failing element ids) |
| Persistence | save_document (save or synchronize with central) |
Parameter access covers instance and type parameters: reads fall back to the element type automatically, and a type write reports how many instances it affects before you trust it. Values are returned as display values in project units. Doors report their room assignment (FromRoom/ToRoom).
Documentation
The full guide with a validated per-tool reference is hosted at
mk92at.github.io/pyrevit-mcp/pyrevit-mcp.html.
It supports English and German (?lang=en|de), light and dark themes, and per-tool anchors
(#find_elements). The web chat UI links into it from the tool reference sidebar.
Requirements
- Autodesk Revit 2024 or newer (developed against Revit 2027) with pyRevit installed
- Python 3.11 or newer and uv
- For the web chat UI: an Anthropic API key
Setup
1. Install the Revit extension. Copy or clone this repository, then add
pyrevit-mcp.extension to pyRevit's extension search paths (pyRevit Settings → Custom
Extension Directories, pointing at this repository's root). Enable the Routes server in
pyRevit's settings. Restart Revit.
2. Verify the bridge. With a model open:
curl http://localhost:48884/pyrevit_mcp/status/
The trailing slash is mandatory on every route.
3. Install the Python side:
uv sync
4a. Register with an MCP client, e.g. Claude Code:
claude mcp add revit -s user -- uv --directory "<path-to>/pyrevit-mcp" run python server.py
4b. Or run the web chat UI. Create a .env with your ANTHROPIC_API_KEY (see
.env.example), then:
uv run uvicorn app:app --port 8100
Open http://localhost:8100. The sidebar has a tool reference generated from the server at
runtime, with a clickable example per tool.
After any change to the extension code, restart Revit cold. The Routes server (Beta)
crashes Revit on a pyRevit reload. scripts/restart_revit.ps1 automates the cold restart and
test_routes.py smoke-tests every route afterwards.
Layout
server.py MCP server, 21 tools, shared HTTP bridge
app.py FastAPI chat backend (CSV, undo, images, downloads)
web/index.html chat UI, single file
test_routes.py end-to-end smoke test for every route
scripts/restart_revit.ps1 automated cold restart
pyrevit-mcp.extension/
startup.py route registration
lib/pyrevit_mcp/
status.py, elements.py, parameters.py, views.py, selection.py, model.py
utils.py ElementId helpers, safe JSON encoding, parameter access
Known limits
- No geometry creation or deletion; this server reads, writes parameters, and visualises.
- Type parameters are written with a warning, but there is no dry-run mode.
- Rooms queries slow down on very large models.
- The Routes server is a pyRevit beta feature; treat the whole bridge as a prototype.
Disclaimer and support
I work at Autodesk as a Technical Specialist for Emerging Tech in Construction. This project grew out of my own exploration of MCP and the Revit API and runs alongside that work, not as part of it. It is not an official Autodesk product or project. Nothing here is built, reviewed, endorsed or maintained by Autodesk, and nothing in this repository is a statement on behalf of Autodesk. Reaching me through my job is not a way to get help with it.
On support I want to be straight with you, because the honest answer has two halves.
I would like to offer as much help as I can, and I am genuinely interested in developing this further together with anyone who finds it useful. Issues and pull requests are welcome and I will read them. Tell me what you built with it.
What I cannot do is promise any of that. There is no guaranteed response, no timeline, no fix for your blocker, and no commitment that this keeps working against future Revit or pyRevit changes. If you are building something that has to work, fork it and own your copy.
This project stands on pyRevit by Ehsan
Iran-Nejad and the pyRevit Labs community, licensed GPL-3.0. pyRevit and its Routes
module are the bridge that makes any of this possible; this repository contains no
pyRevit code and you install pyRevit yourself. Patterns for the ElementId handling and
the route registration were adapted from
revit-mcp-server by Talal Ahmed
(Demolinator), licensed MIT. The code in this repository is my own and is licensed MIT
as well, see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.