Integration Beta

VitruAI + MCP

VitruAI ships an MCP server that exposes Revit, ArchiCAD, AutoCAD, and Rhino to any MCP-compatible AI assistant under the open Model Context Protocol introduced in November 2024. Teams running Claude Desktop, Claude Code, ChatGPT, Codex CLI, Cursor, Windsurf, Zed, or Sourcegraph connect once and drive their BIM tooling with the assistant they already use, with the `claude archicad mcp` workflow currently shipping under Beta with 1–3 design partners.

  • One MCP server exposes Revit, ArchiCAD, AutoCAD, and Rhino through a single integration instead of four separate assistant-specific stacks.
  • Works with the AI assistant your firm already uses — Claude Desktop, Claude Code, ChatGPT (Apps SDK / Developer Mode MCP connectors), Codex CLI, Cursor, Windsurf, Zed, Sourcegraph, and a growing list.
  • Keeps AI-assistant choice in the firm’s hands; VitruAI’s MCP server does not lock you into one vendor or client application.
Apply to the beta See capabilities ↓
Capabilities

Compatibility matrix and what runs through MCP

  • MCP-compatible AI assistants — the canonical list as of 2026-04-29

    VitruAI’s MCP server accepts connections from Claude Desktop, Claude Code, ChatGPT via Apps SDK or Developer Mode MCP connectors, Codex CLI, Cursor, Windsurf, Zed, and Sourcegraph. Any of these clients can call the same BIM-driving tools without separate plug-ins per assistant. As new MCP clients appear, they attach through the same protocol surface without changes to the VitruAI deployment.

  • BIM tools exposed through the MCP server

    The MCP server fronts the same integrations that power VitruAI’s native BIM stack: Revit via the signed add-in described in VitruAI + Revit, ArchiCAD via the VitruAI add-on and Tapir bridge described in VitruAI + ArchiCAD, AutoCAD through the VitruAI plug-in in VitruAI + AutoCAD, and Rhino / Grasshopper through the plug-in in VitruAI + Rhino / Grasshopper. MCP adds the assistant-facing surface on top of these tool-side integrations.

  • Agents callable through MCP

    Once connected, any Live or Beta VitruAI agent becomes callable from the MCP client. That includes the Code Compliance Agent for rule-by-rule checks, the Studio QA/QC Agent for drawing and model review, the Rhino↔Revit Interop Agent for geometry transfer, Document AI for sheet understanding, BOQ/Take-off agents for quantities, and Project Memory for decision logs. The same agent catalog appears in Claude Desktop, Cursor, or any other MCP client without separate wiring.

  • One workspace, one assistant, one BIM tool

    The MCP server authenticates against a VitruAI workspace, then binds that workspace to one or more BIM tools such as Revit or ArchiCAD. The AI assistant connects using the standard MCP local-stdio transport on a developer workstation or remote-SSE from a central server. Within that session, the assistant can read sheets, place views, or run compliance checks as if it were sitting inside the native add-in UI, but under the firm’s existing identity and permission model.

  • The `claude archicad mcp` pattern

    The `claude archicad mcp` configuration — Claude Desktop plus ArchiCAD via MCP — is one of the most common starting points for Beta partners. Tapir bridges the ArchiCAD JSON Command API into a stable HTTP surface; VitruAI’s MCP server then exposes that surface as tools and resources that Claude Desktop or Claude Code can call. The same pattern extends to Revit, AutoCAD, or Rhino once their respective VitruAI integrations are installed in the project environment.

  • Network profile and security envelope

    VitruAI’s MCP server initiates outbound HTTPS-only traffic to a single fully qualified domain name and documented IP range, which simplifies firewall rules for IT. Tenant regions are currently eu-central or me-central, with data residency pinned per tenant. MCP clients on developer laptops or VDI sessions connect to the local MCP server process, which then forwards only the minimum context required for each tool call or document read.

Claude, ArchiCAD, MCP — and every other compatible AI assistant

The `claude archicad mcp` workflow is a concrete example, not a special case. A project architect can open ArchiCAD with the VitruAI add-on installed, start Claude Desktop with the VitruAI MCP server configured, and then ask Claude to run a code check, update a view set, or push geometry to Revit through the same MCP tools. The same MCP endpoint also serves Cursor for Dynamo or pyRevit script editing against VitruAI + Revit projects, or Windsurf and Zed for AutoCAD and Rhino automation.

Common questions

MCP integration — common questions

  • What is MCP?

    MCP is the open Model Context Protocol introduced in November 2024 that defines how AI assistants call tools and read context over a standard JSON-based interface. Clients such as Claude Desktop, Cursor, Windsurf, Zed, ChatGPT via Apps SDK, and Sourcegraph implement this protocol so they can talk to external servers like VitruAI’s MCP endpoint. For BIM teams, that means a single protocol for driving Revit, ArchiCAD, AutoCAD, or Rhino from whichever assistant the firm standardises on.

  • Does this work with Claude, ArchiCAD, MCP — the `claude archicad mcp` pattern specifically?

    Yes, the `claude archicad mcp` pattern is one of the primary Beta configurations. Tapir bridges the ArchiCAD JSON Command API into a service that VitruAI’s MCP server exposes as tools and resources. Claude Desktop or Claude Code then connect to that MCP server, so a designer can ask for a code check, view update, or element query in ArchiCAD using the same commands they would use for Revit or AutoCAD projects.

  • Do we have to use Claude?

    No, VitruAI’s MCP server works with any MCP-compatible AI assistant. Firms can standardise on ChatGPT via Apps SDK or Developer Mode MCP connectors, Codex CLI for terminal-driven workflows, Cursor for code-centric sessions, Windsurf or Zed for IDE-based work, or Sourcegraph for repository-focused reviews. The protocol surface stays the same, so switching assistants later does not require changes to the Revit, ArchiCAD, AutoCAD, or Rhino integrations.

  • Is this an alternative to the Revit, ArchiCAD, or AutoCAD integrations or in addition to them?

    MCP sits in addition to the native BIM integrations, not instead of them. The VitruAI add-ins for Revit, described in VitruAI + Revit, for ArchiCAD in VitruAI + ArchiCAD, for AutoCAD in VitruAI + AutoCAD, and for Rhino / Grasshopper in VitruAI + Rhino / Grasshopper handle the tool-side connectivity. The MCP server adds an assistant-side surface so clients like Claude Desktop or Cursor can call the same agents that the in-app panels use.

  • What about future MCP clients?

    VitruAI integrates against the MCP specification rather than against individual desktop clients. When new assistants adopt MCP, they can connect to the existing VitruAI MCP server using the same configuration pattern that Claude Desktop or ChatGPT Apps SDK use. The compatibility list dated 2026-04-29 is the canonical snapshot for launch, but the expectation is that additional IDEs and chat clients join that list without requiring changes to customer deployments.

  • What does the Beta cohort look like?

    The Beta cohort consists of 1–3 design partners running an MCP-compatible AI assistant alongside their existing BIM stack. Typical patterns include Claude Desktop or Cursor driving Revit models with the Code Compliance Agent and Studio QA/QC Agent, and ArchiCAD teams using the Tapir bridge described in VitruAI + ArchiCAD. Cohort firms work directly with VitruAI’s engineering team on configuration, build-pack design, and prioritisation of additional MCP-exposed agents.

More from VitruAI

Related

Adjacent agents, use cases, integrations, and regulations that pair with this one.

Agent VitruAI Labs

Structural Sizing Agent — preliminary member sizing from the architectural model

The Structural Sizing Agent reads the architectural Revit model, applies the firm’s preliminary-sizing rules, and emits concept-stage member sizes — beams, columns, slabs — for early…

IFCRevitGlobal
Agent VitruAI Labs

MEP Routing Agent — AI for MEP design review

The MEP Routing Agent is VitruAI’s ai for mep design review — it reads a Revit MEP model and evaluates routing decisions against the firm’s design…

RevitGlobal
Agent VitruAI Labs

Document AI Agent — ai document parsing for aec drawings, PDFs, and DWGs

The Document AI Agent extracts structured data — door and window schedules, mechanical components and dimensions, RFI responses, submittal answers, and plan-archive search results — from…

AutoCADIFCRevitGlobalMENA
Agent Live

Comms Agent — AEC project comms agent for meetings to action items

The Comms Agent reads project-meeting transcripts from Zoom, Microsoft Teams, Google Meet, and Slack huddles, extracts decisions and action items with assigned owner and due date,…

Global
Agent Beta

Project Memory Agent — ai project memory for aec, claim-ready timelines on demand

The Project Memory Agent ingests project correspondence, meeting minutes, RFIs, design submissions, change orders, and contract documents into a structured project memory, then produces decision logs,…

BIM 360 / ACCProcoreGlobal
Agent Roadmap

RFI Agent — ai rfi drafting and tracking for aec

The RFI Agent drafts RFIs grounded in the project record—drawings, specifications, prior RFIs, and meeting minutes—and tracks each item’s schedule, scope, and cost implications inside Procore…

BIM 360 / ACCProcoreGlobal
Agent Beta

Contract Agent — ai contract clause monitoring for aec

The Contract Agent reads signed project contracts—owner-architect, design-build, owner-contractor, and subconsultant agreements—and monitors insurance, indemnity, and payment-terms clauses across the project lifecycle. It tracks which clauses…

Global
Agent Beta

Scope Agent — ai scope drift detection for aec projects

The Scope Agent reads the signed contract scope of services and watches RFIs, meeting minutes, design submissions, and email correspondence for asks that fall outside that…

ProcoreGlobal
Next step

Want early access?

Apply to the beta