Control Revit with Plain English

Automate mundane tasks, generate complex geometries, and draft entire floor plans just by talking to your AI assistant. Transform how you build the world with natural language prompts.

Download Toolkit Watch Demo
AI Architecture Representation

Get the Tools

All these tools grant AI agents access to Revit—either by placing the plugin in your Addins directory, or by using the standalone installer.

Revit Add-in

ArchitectureAgent Plugin

The ultimate extension for Autodesk Revit. Includes the intelligent element filter, auto-dimensioning, and room-tagging AI.

  • ✓ Seamless MCP Integration
  • ✓ Automated Floor Plans
  • ✓ Smart Dimensioning
Download Plugin (.ZIP)
v2.4.1 | Requires Revit 2027
Standalone

Hoverboard Studio

A standalone executable dashboard for managing your AI architectural agents, generating scripts, and visualizing workflows.

  • ✓ Requires Revit Integration
  • ✓ Local LLM Integration
  • ✓ Multi-Project Dashboard
Download Installer (.EXE)
v0.1.0 | Windows 10/11
Server Binary

ArchitectureAgent MCP

The standalone MCP server binary. Use this to integrate ArchitectureAgent's capabilities into external AI assistants like Antigravity or Claude.

  • ✓ Universal Protocol
  • ✓ Lightweight & Fast
  • ✓ Direct AI Integration
Download Binary (.EXE)
v1.0.0 | Windows 10/11

How to install & setup MCP Server

Setup in Antigravity

  1. Download the MCP Server Binary above.
  2. Open Antigravity Settings > MCP Servers.
  3. Click "Add New Server" and name it revit-server.
  4. Set the execution path to the downloaded binary location.
  5. Save and restart Antigravity to enable ArchitectureAgent commands.

Setup in Claude Desktop

  1. Download the MCP Server Binary above.
  2. Open your claude_desktop_config.json file.
  3. Add the server under the mcpServers block: 
    "revit-server": {
  "command": "path/to/archai-mcp-server.exe",
  "args": []
}
  4. Restart Claude Desktop for the tools to load.

Using Hoverboard

Discover the powerful features of your AI architectural agents.

🔌

Installation & Connection

Install Hoverboard and the Revit Plugin. Launch both and ensure the connection indicator in the Revit ribbon turns green, confirming seamless communication.

📋

Execution Plans

Before any action is taken in Revit, Hoverboard will generate an Execution Plan. Review this step-by-step logic to understand the AI's intent and approve or modify it.

🌊

Agent Stream

Watch the Agent Stream panel for real-time insights into the AI's thought process, tool executions, and API feedback during complex generation tasks.

✨

Using Skills

Save reusable commands and generation presets as Skills. Hoverboard's AI will automatically select and apply the right skills for your prompt, maintaining consistency and speeding up repetitive tasks.

💬

Managing Chats

Use the chat interface to maintain persistent context. Switch between multiple active sessions for different buildings or floors without losing your project history.

🛠️

Powerful MCP Tools

Leverage impressive tools like Dynamic C# Execution, Procedural Packing Algorithms, and Intelligent Element Filtering directly from your prompt.

Setup Walkthrough

A comprehensive guide to installing Hoverboard, connecting it to Revit, and configuring your intelligence providers for AI-assisted architecture.

1

Installing the Revit Add-in

You can either use the automated installer or download the Revit plugin and MCP server directly from our downloads section.

Automated Setup

When running the Hoverboard Studio or plugin installer, simply specify the correct Revit Add-ins folder when prompted:

%appdata%\Autodesk\Revit\Addins\2027

Manual Setup

  1. Extract the downloaded ArchitectureAgent Plugin manually into the Add-ins directory (%appdata%\Autodesk\Revit\Addins\2027).
  2. Download the ArchitectureAgent MCP Server Binary and save it locally.
  3. Point your intelligence client (e.g., Claude Desktop or Antigravity) to this `.exe` file to register the tools.

Hoverboard vs. Standard MCP Clients

While you are completely free to use the ArchitectureAgent plugin and MCP server directly with any compatible client (like Claude Desktop or Antigravity), Hoverboard Studio comes pre-packaged with a robust library of built-in architectural skills. This makes Hoverboard significantly better at executing complex architectural tasks out-of-the-box, though you are completely free to copy and port these skills to your client of choice!

* Note: ArchitectureAgent requires Revit 2027. Placing the plugin in the Add-ins directory allows Revit to automatically connect upon startup.

2

Verifying the Connection

After the installation is complete, it's time to verify that Revit and Hoverboard are communicating properly.

  1. Launch Autodesk Revit and open any project (or start a new one).
  2. Look for a small notification pop-up in Revit confirming that ArchitectureAgent has initialized successfully.
  3. Launch Hoverboard Studio.

In the Hoverboard interface, check the agent status panel or top navigation bar. The connection indicator light should turn Green. This confirms that the MCP Server has established a real-time, two-way connection with your active Revit session.

3

Setting Up Intelligence Providers

Hoverboard acts as the bridge between Revit and large language models (LLMs). To process your prompts, you must configure an intelligence provider. You can choose a cloud provider or run everything entirely offline.

☁️ Cloud Providers

Use powerful models like Gemini, OpenAI, or Claude via API keys.

  1. Obtain an API key from your preferred provider's developer dashboard.
  2. In Hoverboard Studio, click the Settings (gear icon).
  3. Navigate to the API Keys section.
  4. Paste your API key into the respective field and click Save.
  5. In the chat interface, select your cloud model from the dropdown to begin.

💻 Local LLMs (Ollama)

Run models locally for ultimate privacy and offline usage.

  1. Download and install Ollama on your machine.
  2. Open your terminal and pull a model:
    ollama run llama3
  3. In Hoverboard Settings, go to Local Models.
  4. Verify the Ollama endpoint is http://localhost:11434
  5. Select the local model in your chat interface to run commands completely offline.