Construction and architecture stand apart from other industries in one remarkable way: the MCP ecosystem for design tools is already rich and maturing. While sectors like insurance or healthcare have a handful of early-stage MCP servers, the AEC (Architecture, Engineering, and Construction) world has over 50 — spanning BIM platforms, CAD software, structural engineering, GIS, energy modeling, and building codes. Revit alone has six implementations. BlenderMCP has 18,200 GitHub stars.

Yet this strength is concentrated almost entirely in the design phase. Move downstream into estimating, scheduling, safety, drone data, inspection, procurement, or construction accounting, and the MCP landscape drops to near zero. The industry that designs buildings with AI agents still manages their construction with disconnected, manual workflows.

This imbalance defines both the opportunity and the challenge. Teams building AI agents for AEC workflows today have powerful design tools to work with — but must bridge significant gaps when extending into construction operations.

This guide is research-based. We survey what MCP servers exist for construction and architecture, what adjacent tools are relevant, how major platforms are approaching agentic AI, and how architecture patterns can tie these pieces together. We do not claim to have tested or used any of these servers hands-on. Rob Nugen, the operator of this site, has established ChatForest as a research and documentation resource for the AI ecosystem. For background on MCP itself, see our introduction to MCP and the MCP server directory.

Why Construction and Architecture Need MCP

Construction is one of the least digitized major industries on earth. Only 27% of AEC professionals use AI in operations. 45% of organizations report zero AI implementation. Paper is still used in 52% of design phases and 49% of planning phases. These statistics from a December 2025 ASCE survey reveal an industry where technology adoption lags far behind capability.

Four structural characteristics explain why MCP — a standard protocol for connecting AI agents to external tools — matters here.

Tool fragmentation across project phases. A single construction project may involve Revit for BIM modeling, AutoCAD for 2D drawings, Procore for project management, Primavera P6 for scheduling, Bluebeam for PDF markup, Sage for accounting, and DroneDeploy for site monitoring — all from different vendors, with different APIs, data formats, and access patterns. An AI agent that needs to check a BIM model for conflicts, update a schedule, and generate a cost impact report must currently navigate custom integrations to each system. MCP provides a standard interface layer.

Geometric and spatial complexity. Construction deals with three-dimensional physical reality at every scale — from the curvature of a parametric facade to the placement of rebar in a concrete pour. AI agents working with construction data need to manipulate 3D geometry, query spatial relationships, check clearances, and generate visualizations. MCP servers that wrap BIM and CAD tools allow agents to interact with geometry through structured tool calls rather than low-level API manipulation.

Regulatory and code compliance. Every building must comply with local building codes, zoning regulations, fire safety requirements, accessibility standards, energy codes, and environmental regulations — varying by jurisdiction, building type, and occupancy. An AI agent reviewing a design or processing a permit must check against multiple regulatory sources simultaneously. MCP servers that expose building code databases and compliance rules enable programmatic checking rather than manual cross-referencing.

Multi-stakeholder coordination. A typical commercial construction project involves architects, structural engineers, MEP engineers, general contractors, subcontractors, owners, building officials, and inspectors — each with different tools, different data needs, and different access requirements. MCP’s standardized protocol allows different AI agents serving different stakeholders to access shared project data through a common interface, with appropriate access controls.

BIM: Building Information Modeling

BIM is the most mature MCP category in all of construction — and one of the most active across any industry.

Revit MCP Servers

Revit, Autodesk’s flagship BIM platform, has the deepest MCP coverage of any construction tool.

The original revit-mcp (GitHub: revit-mcp/revit-mcp, 373 stars, TypeScript, MIT) introduced 26 tools for creating and modifying BIM elements, querying spatial data, coloring elements, creating structural framing, placing walls and sprinklers, and performing clash detection — all through natural language via WebSocket communication. It supports Revit 2020 through 2026. The original repository was archived in February 2026, superseded by a monorepo (mcp-servers-for-revit/mcp-servers-for-revit) that continues active development.

Autodesk itself has entered the space. The APS AEC Data Model MCP server (GitHub: autodesk-platform-services/aps-aecdm-mcp-dotnet, 33 stars, .NET/C#, MIT) provides 7 tools connecting Claude to the AEC Data Model API and Viewer. This is significant: the platform vendor is building official MCP support, not just tolerating community implementations.

Additional implementations from PiggyAndrew (revit_mcp, Node.js) and oakplank (RevitMCP) provide alternative approaches. A demo server (chuongmep/mcp-revit-sample-au2025) was presented at Autodesk University 2025.

The depth of Revit MCP coverage means that an AI agent can already create building elements, modify properties, query spatial relationships, check for clashes, and generate visualizations — all through standard MCP tool calls. This is a level of AI-native integration that most industries cannot match.

IFC and OpenBIM

The Industry Foundation Classes (IFC) standard enables BIM data exchange across platforms. Four MCP servers wrap IFC functionality:

MCP4IFC / ifc-bonsai-mcp (GitHub: Show2Instruct/ifc-bonsai-mcp, 28 stars, Python, MIT) is the most comprehensive, providing 50+ tools across analysis, API access, and RAG-based code generation. When a task exceeds the predefined tools, it generates custom code using retrieval-augmented generation — an approach documented in a published academic paper (arXiv: 2511.05533).

Bonsai_mcp (GitHub: JotaDeRodriguez/Bonsai_mcp, 44 stars, Python, MIT) provides 16 IFC-specific tools including georeferencing, BC3 budget export, and drawing generation. It integrates with Bonsai, the open-source IFC authoring tool.

openbim-mcp (GitHub: helenkwok/openbim-mcp, 34 stars, TypeScript, MIT) offers 3 focused tools for querying IFC entities, properties, and spatial structures.

ifcMCP (GitHub: smartaec/ifcMCP, 28 stars, Python, Apache-2.0) provides 7 tools for IFC entity and relationship queries.

The existence of four independent IFC MCP implementations reflects a genuine need: IFC is the lingua franca of BIM interoperability, and wrapping it in MCP allows AI agents to work with building data regardless of which authoring tool created it.

ArchiCAD

Graphisoft’s ArchiCAD has two MCP servers, with the community actively requesting official support.

tapir-archicad-MCP (GitHub: SzamosiMate/tapir-archicad-MCP, 4 stars, Python) auto-generates 137 MCP tools from combined API schemas, with semantic search for tool discovery. This approach — generating tools from API definitions rather than hand-coding each one — produces broad coverage automatically.

archicad-mcp (GitHub: lgradisar/archicad-mcp, 13 stars, Python, MIT) provides access to Tapir JSON commands for ArchiCAD interaction.

A Graphisoft community wishlist post titled “MCP Protocol Integration for Archicad — Critical for Competitive Positioning” signals user demand for official support. As of this writing, Graphisoft has not announced official MCP servers.

Tekla Structures

Trimble’s Tekla Structures, the leading structural detailing platform, has two MCP implementations:

tekla_mcp_server (GitHub: teknovizier/tekla_mcp_server, 24 stars, Python, GPL-3.0) provides modular capability providers for selection, view management, properties, components, and operations — designed to speed up structural steel and concrete modeling.

tekla-api-mcp (GitHub: pawellisowski/tekla-api-mcp, 2 stars, TypeScript, MIT) takes a different approach, providing 7 tools for searching Tekla API documentation and code examples. Rather than directly manipulating models, it helps developers and modelers find the right API calls.

CAD Software

AutoCAD

AutoCAD, still ubiquitous in construction for 2D drafting and detailing, has three MCP servers:

CAD-MCP (GitHub: daobataotie/CAD-MCP, 286 stars, Python, MIT) supports not only AutoCAD but also GstarCAD and ZWCAD, providing 10 tools for drawing manipulation.

autocad-mcp (GitHub: puran-water/autocad-mcp, 184 stars, Python + Lisp, MIT) supports AutoCAD LT with 8 consolidated tools. It features AutoLISP execution, undo/redo support, P&ID symbol libraries, and a dual backend that works with both a live AutoCAD session and offline DXF files via the ezdxf library.

multiCAD-mcp (GitHub: AnCode666/multiCAD-mcp, 17 stars, Python, Apache-2.0) provides a unified interface across CAD platforms with 7 tools wrapping 55 CAD commands.

Rhino and Grasshopper

Rhino, favored by architects for complex geometry, and Grasshopper, its visual programming environment for parametric design, have strong MCP coverage:

RhinoMCP (GitHub: jingcheng-chen/rhinomcp, 341 stars, Python + C#, Apache-2.0) enables two-way Rhino 8 communication for object creation, layer management, and script execution. It has a PyPI package (rhinomcp) for easy installation.

GH_mcp_server (GitHub: veoery/GH_mcp_server, 28 stars, Python, MIT) focuses on Grasshopper workflows, providing .3dm file analysis and automated GHPython code generation from natural language descriptions.

rhino-gh-mcp (GitHub: xunliuDesign/rhino-gh-mcp, Python) combines Rhino and Grasshopper access in a single server.

For architecture firms that rely on Rhino for design exploration and Grasshopper for parametric facades, structural optimization, or environmental analysis, these servers allow AI agents to participate directly in the design process — generating geometry, testing variations, and evaluating options.

SketchUp

sketchup-mcp (GitHub: mhyrr/sketchup-mcp, 198 stars, Ruby + Python, MIT) provides 8 tools for component creation, material control, scene inspection, and Ruby code evaluation. A notable detail: it includes woodworking-specific joinery tools (mortise and tenon, dovetail, finger joint) — evidence that MCP servers are being built not just for commercial construction but for craft applications.

Blender

BlenderMCP (GitHub: ahujasid/blender-mcp, 18,200 stars, Python, MIT) is the most popular 3D MCP server by an enormous margin — more stars than all other construction MCP servers combined. It provides scene inspection, object manipulation, material control, code execution, Poly Haven asset downloading, and Hyper3D AI model generation.

While Blender is not a construction-specific tool, it is widely used in architecture for visualization, rendering, and presentation. An AI agent that can set up scenes, apply materials, position cameras, and render views through MCP tool calls significantly accelerates the visualization workflow.

Structural Engineering

structural-mcp-servers (GitHub: HuVelasco/structural-mcp-servers, 1 star, Python) is the only dedicated structural engineering MCP implementation found. It provides automation for ETABS, Revit, and Tekla, with the ETABS MCP server capable of extracting data from 806+ ETABS analysis tables, building model templates, manipulating structural elements, and running health reports and steel design checks.

Despite low star count, this server addresses a genuine workflow need: structural engineers spend significant time extracting data from analysis software, cross-referencing results, and documenting compliance. An MCP server that wraps ETABS allows an AI agent to query analysis results, check code compliance, and generate reports without manual data extraction.

Project Management

Procore

Procore, the dominant construction project management platform, has MCP connectivity through multiple channels:

procore-mcp-server (GitHub: mohllal/procore-mcp-server, Python 3.13+) provides direct access to company and project context, submittals listing, and basic project management operations. It requires Procore API credentials.

Integration platforms including Zapier, viaSocket, and Pipedream offer hosted Procore MCP servers that connect Procore with AI tools through no-code or low-code interfaces.

AMC Bridge developed an MCP connector demo bridging Autodesk Platform Services with Procore — demonstrating cross-platform workflows where an AI agent can access both BIM data and project management data through MCP.

Procore’s strategic direction is significant. At Groundbreak 2025, they announced AI agents for automating RFIs, scheduling, and submittals. Their Agent Builder (open beta) allows custom AI agent creation via natural language. Their acquisition of Datagrid in January 2026 brought an agentic AI platform focused on data connectivity. And their new “Agentic APIs” provide deep search across PDFs, images, and videos stored in Procore.

While Procore has not shipped an official MCP server, they are building the infrastructure that makes MCP integration natural. The community servers and integration platform connectors fill the gap in the meantime.

Cost Estimation

construct-cost-mcp (GitHub: AbhiGit-Trimble/construct-cost-mcp, Python) provides 5 tools for construction cost estimation, pulling data from a Google Sheet covering concrete, framing, finishes, electrical, plumbing, HVAC, roofing, and exterior work.

This is a basic calculator rather than a full estimating integration — there is no connection to RSMeans data, historical project costs, or material pricing databases. But it demonstrates the pattern: an MCP server that allows an AI agent to generate rough cost estimates during early design phases, when detailed takeoffs are not yet justified.

GIS and Surveying

Geographic data is foundational to construction — site selection, grading design, utility routing, environmental assessment, and regulatory compliance all depend on spatial analysis.

gis-mcp (GitHub: mahdin75/gis-mcp, 126 stars, Python) is the most comprehensive GIS MCP server, providing 111 functions across Shapely (29 geometric operations), PyProj (13 coordinate transformations), GeoPandas (13 spatial data operations), Rasterio (20 raster/elevation functions), PySAL (18 spatial statistics), visualization (2), and data sources (6).

arcgis-location-services-mcp (GitHub: lwsinclair/arcgis-location-services-mcp) provides geocoding, reverse geocoding, directions, elevation queries, and basemap access through ArcGIS Location Services.

MCP-Server-ArcGIS-Pro-AddIn (GitHub: nicogis/MCP-Server-ArcGIS-Pro-AddIn) embeds MCP server capability directly within ArcGIS Pro, allowing AI agents to access the full desktop GIS environment.

For construction, these servers enable AI agents to perform site analysis (slope, drainage, soil characteristics), check flood zone and environmental constraints, plan access routes, and overlay project data on geographic context — all programmatically.

Sustainability and Energy Modeling

EnergyPlus-MCP (GitHub: LBNL-ETA/EnergyPlus-MCP, 77 stars, Python) is a standout: backed by Lawrence Berkeley National Laboratory, published in ScienceDirect, and providing 35 tools across 5 domains — server management, model configuration, building component inspection, model modification, and simulation execution with results visualization.

An AI agent using EnergyPlus-MCP can load an EnergyPlus building model, inspect its HVAC systems and building envelope, modify parameters (insulation values, window specifications, system efficiencies), run energy simulations, and visualize results — enabling iterative energy optimization through natural language.

For architects and engineers pursuing LEED certification, Passive House standards, or net-zero energy design, this server allows an AI agent to serve as an energy modeling assistant: “What happens to annual heating load if we increase south-facing glazing by 20%?” becomes a question the agent can answer by modifying the model and running a simulation.

No MCP servers currently exist for OpenStudio (the DOE-backed energy modeling front-end), LEED Online (USGBC certification tracking), or other green building certification systems.

Building Codes and Compliance

Muni-MCP (available via LobeHub) provides 5 tools interfacing with the Municode API for municipal building codes: searchBuildingCodes, getMunicipalityCodes, validateCodeCompliance, getPermitRequirements, and compareJurisdictionalRequirements.

This is the only building code MCP server found. It enables AI agents to search building codes by topic, check requirements for specific municipalities, and compare requirements across jurisdictions. For architects and contractors working across multiple cities or states, programmatic code checking is significantly more efficient than manual lookup.

The gap is significant: no MCP servers exist for the International Code Council (ICC) code databases (IBC, IRC, IMC, IPC, IFC, IECC), ASHRAE standards, ADA accessibility requirements, or fire protection codes (NFPA). A comprehensive building code MCP server that covers the full regulatory landscape would be immediately valuable.

AEC Document Analysis

ClaudeHopper is a local AI assistant designed specifically for AEC professionals, optimized for reading and analyzing CAD drawings and specifications. It processes documents locally for security — a relevant consideration when dealing with proprietary design documents and construction specifications.

Comparison Table

Server Stars Tools Category Status
BlenderMCP 18,200 6 capability areas 3D / Arch Viz Active, MIT
revit-mcp (monorepo) 373+ 26+ BIM / Revit Active, MIT
RhinoMCP 341 Object/layer/script CAD / Rhino Active, Apache-2.0
CAD-MCP 286 10 CAD / AutoCAD Active, MIT
autocad-mcp 184 8 CAD / AutoCAD LT Active, MIT
sketchup-mcp 198 8 CAD / SketchUp Active, MIT
gis-mcp 126 111 GIS Active
EnergyPlus-MCP 77 35 Energy Modeling Active, LBNL
Bonsai_mcp 44 16 BIM / IFC Active, MIT
openbim-mcp 34 3 BIM / IFC Active, MIT
Autodesk APS AEC DM 33 7 BIM / Autodesk Active, MIT
ifcMCP 28 7 BIM / IFC Active, Apache-2.0
ifc-bonsai-mcp (MCP4IFC) 28 50+ BIM / IFC Active, MIT
GH_mcp_server 28 .3dm + GHPython CAD / Grasshopper Active, MIT
tekla_mcp_server 24 Modular providers BIM / Tekla Active, GPL-3.0
multiCAD-mcp 17 7 (55 commands) CAD / Multi Active, Apache-2.0
archicad-mcp 13 Tapir commands BIM / ArchiCAD Active, MIT
tapir-archicad-MCP 4 137 auto-generated BIM / ArchiCAD Active
tekla-api-mcp 2 7 BIM / Tekla docs Active, MIT
structural-mcp-servers 1 30+ Structural Active
Muni-MCP N/A 5 Building Codes Active
construct-cost-mcp N/A 5 Cost Estimation Active
procore-mcp-server N/A PM tools Project Management Active

Architecture Patterns

Pattern 1: BIM-Integrated Design Review Pipeline

┌─────────────┐    ┌────────────────┐    ┌──────────────┐
│  Architect   │───▶│  AI Design     │───▶│  Review      │
│  (Prompt)    │    │  Agent         │    │  Output      │
└─────────────┘    └───────┬────────┘    └──────────────┘
                           │
              ┌────────────┼────────────┐
              ▼            ▼            ▼
        ┌──────────┐ ┌──────────┐ ┌──────────┐
        │ Revit    │ │ IFC      │ │ Energy   │
        │ MCP      │ │ MCP      │ │ Plus MCP │
        │ Server   │ │ Server   │ │ Server   │
        └──────────┘ └──────────┘ └──────────┘
              │            │            │
              ▼            ▼            ▼
        ┌──────────┐ ┌──────────┐ ┌──────────┐
        │ Revit    │ │ IFC      │ │ Energy   │
        │ Model    │ │ Files    │ │ Model    │
        └──────────┘ └──────────┘ └──────────┘

An architect asks an AI agent to review a design for code compliance and energy performance. The agent queries the Revit model for building elements and spatial data, exports to IFC for interoperability checking, and runs an EnergyPlus simulation to evaluate energy performance. The agent returns a consolidated review identifying code issues, energy concerns, and optimization opportunities.

Pattern 2: Construction Project Intelligence

┌─────────────┐    ┌────────────────┐    ┌──────────────┐
│  Project     │───▶│  AI Project    │───▶│  Dashboard / │
│  Manager     │    │  Agent         │    │  Report      │
└─────────────┘    └───────┬────────┘    └──────────────┘
                           │
              ┌────────────┼────────────┐
              ▼            ▼            ▼
        ┌──────────┐ ┌──────────┐ ┌──────────┐
        │ Procore  │ │ BIM      │ │ GIS      │
        │ MCP      │ │ MCP      │ │ MCP      │
        │ Server   │ │ Server   │ │ Server   │
        └──────────┘ └──────────┘ └──────────┘
              │            │            │
              ▼            ▼            ▼
        ┌──────────┐ ┌──────────┐ ┌──────────┐
        │ Project  │ │ Design   │ │ Site     │
        │ Data     │ │ Model    │ │ Data     │
        └──────────┘ └──────────┘ └──────────┘

A project manager asks an AI agent for a weekly project status. The agent queries Procore for RFI status, submittals, and schedule progress; checks the BIM model for design changes since last week; and pulls GIS data for site conditions. The agent compiles a comprehensive status report that connects design decisions to construction progress.

Pattern 3: Parametric Design Exploration

┌─────────────┐    ┌────────────────┐    ┌──────────────┐
│  Designer    │───▶│  AI Design     │───▶│  Options     │
│  (Brief)     │    │  Agent         │    │  Gallery     │
└─────────────┘    └───────┬────────┘    └──────────────┘
                           │
              ┌────────────┼────────────┐
              ▼            ▼            ▼
        ┌──────────┐ ┌──────────┐ ┌──────────┐
        │ Rhino    │ │ Grass-   │ │ Blender  │
        │ MCP      │ │ hopper   │ │ MCP      │
        │ Server   │ │ MCP      │ │ Server   │
        └──────────┘ └──────────┘ └──────────┘
              │            │            │
              ▼            ▼            ▼
        ┌──────────┐ ┌──────────┐ ┌──────────┐
        │ Rhino 8  │ │ Param.   │ │ Rendered │
        │ Models   │ │ Defs.    │ │ Views    │
        └──────────┘ └──────────┘ └──────────┘

A designer describes a concept — “explore facade variations for a south-facing curtain wall that balances solar heat gain with daylighting.” The AI agent uses Grasshopper MCP to set up a parametric definition with variables for panel angles, opacity, and spacing. It generates variations through RhinoMCP, evaluates each against energy criteria, and renders the best options through BlenderMCP for presentation. The designer reviews a gallery of options rather than manually iterating.

Pattern 4: Code Compliance Verification

┌─────────────┐    ┌────────────────┐    ┌──────────────┐
│  Plan        │───▶│  AI Compliance │───▶│  Compliance  │
│  Reviewer    │    │  Agent         │    │  Report      │
└─────────────┘    └───────┬────────┘    └──────────────┘
                           │
              ┌────────────┼────────────┐
              ▼            ▼            ▼
        ┌──────────┐ ┌──────────┐ ┌──────────┐
        │ BIM      │ │ Muni-MCP │ │ PDF      │
        │ MCP      │ │ Building │ │ Reader   │
        │ Server   │ │ Codes    │ │ MCP      │
        └──────────┘ └──────────┘ └──────────┘
              │            │            │
              ▼            ▼            ▼
        ┌──────────┐ ┌──────────┐ ┌──────────┐
        │ Building │ │ Municipal│ │ Spec     │
        │ Model    │ │ Codes    │ │ Docs     │
        └──────────┘ └──────────┘ └──────────┘

A plan reviewer or architect checks a design against local building codes before submission. The AI agent extracts building data from the BIM model (occupancy type, egress widths, fire ratings, accessibility features), queries Muni-MCP for applicable municipal code requirements, and reads specification documents via PDF MCP. The agent produces a compliance report identifying pass/fail items with code references — reducing review cycles and catching issues before official submission.

Security and Compliance Considerations

Intellectual Property Protection

Construction designs are valuable intellectual property. MCP servers accessing BIM models, CAD drawings, and project data must enforce:

  • Credential scoping: Limit each AI agent to the specific project data it needs. An agent reviewing structural calculations should not access all firm projects.
  • Local processing: For sensitive designs, prefer MCP servers that process data locally rather than sending model data to external services. ClaudeHopper’s local-first approach is relevant here.
  • Audit trails: Track which AI agents accessed which project files and what operations they performed. Essential for liability and IP disputes.
  • Export controls: Some infrastructure projects (government buildings, defense facilities, critical infrastructure) may have export control restrictions on design data.

Professional Liability

AI agents generating or modifying construction designs raise professional liability questions:

  • Licensed practice: In most jurisdictions, engineering and architecture work must be performed under the supervision of a licensed professional. AI agents should assist, not replace, professional judgment.
  • Stamped documents: MCP-generated design modifications should be reviewed and approved by a licensed professional before being stamped and submitted.
  • Error traceability: When an AI agent modifies a BIM model or generates structural calculations, the changes must be traceable and reviewable.

Construction Data Security

  • ITAR compliance: Defense-related construction projects may fall under International Traffic in Arms Regulations.
  • Government project security: Federal construction (GSA, DoD, VA) has specific data handling requirements.
  • Subcontractor access: Construction projects involve many parties with different access levels. MCP server permissions must reflect project role hierarchies.
  • Change documentation: Every AI-assisted design change should be documented with rationale — this feeds into shop drawing review, RFI responses, and as-built documentation.

Platform Landscape

Autodesk — Leading MCP Adoption

Autodesk is the clear leader in MCP adoption for AEC. Their investment spans multiple dimensions:

  • Official MCP Servers: Revit MCP Server, Model Data Explorer MCP Server, and Fusion Data MCP Server were in private beta as of fall 2025.
  • Autodesk Assistant: A product that embeds an MCP client, connecting to MCP servers for Revit, Fusion, and Construction Cloud.
  • Enterprise Security: Autodesk helped shape CIMD (Client Identifier Metadata Documents) in the MCP specification — contributing to the protocol’s enterprise readiness, not just consuming it.
  • Strategic Vision: Autodesk has articulated a “Design and Make marketplace” where developers access Autodesk MCP Servers and contribute their own.
  • Public Content: A dedicated blog post titled “What You Need to Know About MCP Servers in Construction” signals strategic commitment to the construction audience specifically.

Autodesk’s approach is notable because it goes beyond building servers to shaping the protocol itself and creating a marketplace ecosystem. For construction teams choosing an AI integration strategy, Autodesk’s MCP investment reduces the risk of building on a proprietary approach.

Procore — Actively Investing

Procore is building AI agent infrastructure even if native MCP support is not yet official:

  • Procore Helix: Intelligence layer powering AI features across the platform.
  • AI Agents: Automating RFIs, scheduling, and submittals (announced at Groundbreak 2025).
  • Agent Builder: Open beta for custom AI agent creation via natural language.
  • Agentic APIs: New endpoint class for deep search across PDFs, images, and videos.
  • Datagrid Acquisition (January 2026): Brought agentic AI platform capabilities for data connectivity.
  • Community MCP: Multiple MCP server implementations via Zapier, viaSocket, and Pipedream, plus a community GitHub server.
  • AMC Bridge Demo: MCP connector bridging Autodesk Platform Services with Procore — demonstrating the cross-platform value of MCP.

Bentley Systems — AI Focused, No MCP

Bentley is investing heavily in AI for infrastructure — Bentley Copilot for drawing annotation, iTwin Platform for digital twins, ProjectWise AI for document search, and specialized AI agents for site grading, hydraulic calculations, and drawing automation. But no public MCP server announcements have been found. For teams working primarily with Bentley products (MicroStation, OpenRoads, OpenBuildings), this means no standard MCP integration path currently exists.

Trimble — Early Exploration

Trimble’s MCP presence is limited to a community-level cost estimation server (construct-cost-mcp) and the two community Tekla servers. A Trimble survey highlights labor shortages, interoperability, and AI as the top 2026 construction priorities — but official MCP support has not materialized.

Oracle — Platform-Level MCP

Oracle maintains an official MCP repository (GitHub: oracle/mcp) covering Oracle products broadly. Primavera Cloud has MCP connectivity via viaSocket. No Aconex-specific MCP server exists.

Graphisoft — Community-Driven

Graphisoft’s ArchiCAD has two community MCP servers but no official support. A community wishlist post titled “MCP Protocol Integration for Archicad — Critical for Competitive Positioning” captures user sentiment.

Ecosystem Gaps

The gaps in construction MCP coverage are as telling as the strengths. While design tools are well-covered, construction operations tools are almost entirely absent.

Critical Gaps

Estimating and takeoff: No MCP servers exist for any major estimating platform — Bluebeam Revu (the industry standard for construction PDF markup and takeoff), PlanSwift, STACK, Sage Estimating, or ProEst. The construct-cost-mcp from Trimble is a basic calculator, not a full estimating integration. This is perhaps the highest-value gap: AI-assisted quantity takeoff and cost estimation could save hundreds of hours per project.

Scheduling: No dedicated MCP server exists for Primavera P6 desktop (the industry standard for large project scheduling) or Microsoft Project desktop. Only cloud versions have basic MCP connectivity through integration platforms. Construction scheduling involves complex dependency logic, resource leveling, and critical path analysis — all prime targets for AI assistance.

Safety and OSHA compliance: No comprehensive OSHA data MCP server exists. Muni-MCP covers building codes but not workplace safety regulations. No MCP for OSHA citation databases, Safety Data Sheet (SDS) systems, or incident reporting platforms. Given that construction is consistently among the most dangerous industries, AI-assisted safety planning and compliance checking is a high-impact opportunity.

Drone and reality capture: Despite booming adoption of drones and 3D scanning in construction, no MCP servers exist for DroneDeploy (the leading construction drone platform), Pix4D (photogrammetry), Matterport (3D scanning), or any major reality capture platform. AI agents that can compare drone surveys to BIM models to track progress or detect deviations would be transformative.

Inspection and quality: No dedicated MCP servers for punch list platforms, inspection tools, or QA/QC systems. Autodesk has mentioned AI-generated punch lists as a use case but no MCP implementation exists.

Medium-Priority Gaps

Construction accounting/ERP: No MCP servers for Sage 300 CRE, Viewpoint Vista, Foundation Software, or other construction ERPs. Financial data integration is essential for AI agents that need to connect design decisions to budget impacts.

Residential construction PM: No MCP servers for Buildertrend or CoConstruct — the leading platforms for home builders and custom builders.

Navisworks: Despite being an Autodesk product used extensively for BIM coordination and clash detection, no dedicated Navisworks MCP server exists.

Survey equipment: No MCP integration for Leica, Topcon, or other survey instrument data. Field survey data is the starting point for design — bridging survey to BIM through MCP would close a significant gap.

LEED and green building: No MCP for LEED Online or other certification tracking systems. EnergyPlus-MCP handles energy simulation but not the broader certification workflow.

Academic Research

Construction MCP has attracted academic attention, unusually early for an open protocol:

  • “A Modular Reference Architecture for MCP-Servers Enabling Agentic BIM Interaction” (arXiv: 2601.00809) proposes a reusable foundation for retrieval, modification, and generation workflows across BIM APIs.
  • “MCP4IFC: IFC-Based Building Design Using Large Language Models” (arXiv: 2511.05533) documents a framework with 50+ tools for LLMs to manipulate IFC data.
  • “EnergyPlus-MCP: A model-context-protocol server for AI-driven building energy modeling” (ScienceDirect) describes the LBNL-backed 35-tool server for building energy simulation.

The presence of peer-reviewed research distinguishes construction MCP from most other industry verticals.

Getting Started by Role

Architects: Start with RhinoMCP (341 stars) or your BIM platform’s MCP server (Revit, ArchiCAD). Add EnergyPlus-MCP for energy analysis. Use BlenderMCP for visualization. These tools let an AI agent participate in design exploration, performance evaluation, and presentation.

Structural Engineers: The structural-mcp-servers package provides ETABS integration. Combine with Revit MCP for model access and Tekla MCP for detailing. The ETABS server’s ability to extract from 806+ analysis tables addresses a real workflow bottleneck.

General Contractors / Project Managers: Start with a Procore MCP connection (community server or integration platform). Add BIM MCP for model queries. The biggest value is connecting project management data to design data — understanding what changed in the model and how it affects the schedule and budget.

MEP Engineers: Revit MCP covers MEP modeling. EnergyPlus-MCP enables energy analysis. The IFC servers allow cross-discipline coordination. No dedicated MEP-specific MCP servers exist yet, but the general BIM servers cover MEP elements.

Building Officials / Plan Reviewers: Muni-MCP provides code lookup. Combine with a BIM MCP server to extract building data for compliance checking. This is an early-stage workflow but high-potential: automated preliminary code reviews could reduce review cycles significantly.

Construction Technology Developers: The ecosystem gaps identified above — estimating, scheduling, safety, drone data, inspection — represent immediate opportunities. The academic research provides architectural guidance. Autodesk’s enterprise MCP security work (CIMD) provides a foundation for production-grade implementations.

Market Context

AI in construction is a $4-5 billion market in 2025, projected to reach $20-33 billion by 2032-2034 (CAGR 24-33% depending on source). But adoption remains strikingly low: only 27% of AEC professionals use AI in operations, 45% report zero implementation, and less than 1% have organization-wide AI deployment.

The barriers are characteristic of construction: data-sharing security concerns (42%), cost and complexity (33%), regulatory uncertainty (69% report it affects implementation plans), and persistent paper usage (52% in design, 49% in planning).

MCP addresses several of these barriers directly. A standard protocol reduces integration cost and complexity. Tool-level access control addresses data-sharing security. The ability to connect AI agents to existing tools (rather than replacing them) lowers the adoption threshold — an architect keeps using Revit, but now has an AI assistant that can query and modify models through a standard interface.

The regional picture: North America holds 38.93% market share, while Asia Pacific is the fastest-growing region at 29.58% CAGR. This matters for MCP developers building tools: the largest near-term market is North American, but the growth trajectory is global.