Revizto Wires ChatGPT and Claude Into Live BIM Data via MCP

Picture a construction site as a busy railway junction. The trains are the specialist tools (BIM platforms, spreadsheets, business intelligence dashboards), and every one of them runs on a different gauge of track. For years the answer has been to build another platform, another siding, another train. Revizto has just done something more interesting: it's laying a single interchange rail that any AI locomotive can pull into.

What Happened

Revizto has launched a Model Context Protocol (MCP) Server, a Model & Object Properties API, and a Developer Portal that together let architecture, engineering, construction and operations teams point ChatGPT, Claude, Microsoft Copilot or any enterprise LLM directly at live Revizto project data. As Construction UK Magazine reported, the capabilities ship under the existing Revizto subscription rather than as a paid add-on.

The framing is deliberate. Revizto's own Bridging the Gap Report 2026 found that lack of time and capacity is the industry's biggest barrier to technology adoption for the second year running, with 32% of construction leaders saying their teams don't have the bandwidth to learn new tools. Meanwhile 96% of organizations flagged growing concerns about data ownership and control, and 39% plan to simplify their technology stack. If you're keeping score, that's three separate signals all pointing at the same fatigue: fewer platforms, tighter governance, less integration work.

Marc Schütz, Revizto's Chief Product Officer, put the pitch bluntly: "We're moving into an era where AI alone won't be the competitive advantage. The organizations that outperform will be those that can securely connect intelligence to trusted, live project data." The MCP Server is the interchange rail. The Properties API is the freight manifest. The Developer Portal, with OAuth 2.0 registration replacing manual access-code handling, is the ticket office. All three land at once, which is the interesting part. Anyone who has shipped a partial integration story to a large enterprise buyer knows the whole thing falls over unless the auth model, the data plane and the developer experience arrive together.

Technical Anatomy

Let's get into the guts of it. MCP (Model Context Protocol) is the emerging standard for exposing structured data and tools to large language models without hand-rolling a bespoke integration for every vendor. Revizto's MCP Server sits between the project data and whichever LLM the customer has already chosen, translating complex project structures into a format an AI agent can query. No custom integration, no DevOps specialist babysitting the pipeline.

The prompts Revizto highlights tell you exactly what job it's designed for: "Which fire doors are missing certification?" or "What are the top five unresolved clashes on Level 3?" These are the boring but expensive questions that normally require a coordinator to open the model, run filters, export a report, and stitch it into a spreadsheet. Collapse that loop into a natural-language round trip and you've killed a category of manual audits.

The Model & Object Properties API is the more analytics-flavoured piece. It exposes real-time model and object metadata (material types, systems, fire ratings, quantities) without opening the Revizto application. That last clause matters more than it sounds. It means the data can be pulled straight into quantity take-offs, cost estimates, live dashboards tracking installation progress and carbon footprints, or piped into procurement and facility management systems for material purchases and digital handover. This is the data-warehouse angle: treat the model as a queryable source, not a static deliverable to be re-exported every time someone asks a question.

For teams already running modern analytics stacks, the pattern is familiar. Pull object metadata via the API, land it in a columnar store like ClickHouse or Snowflake, model it in dbt, then hand the semantic layer to the LLM through MCP. The Developer Portal's OAuth 2.0 flow is what makes this palatable to a security review. Manual access codes are how integrations quietly rot; standards-based auth is how they survive audits.

Who Gets Burned

The obvious losers are the middleware vendors whose entire pitch was "we'll integrate your BIM data with your BI tool." If the primary platform ships a first-party API and an MCP Server under the existing subscription, the case for a bespoke connector layer collapses. That 39% of organizations planning to simplify their tech stack? They're the buying committee, and they've already decided.

Traditional BI dashboards built on nightly extracts also look shaky. When a project manager can ask an LLM "top five unresolved clashes on Level 3" and get an answer in seconds, the value of a static Power BI report that refreshes every twelve hours drops significantly. The dashboard doesn't die, it just becomes the fallback rather than the primary interface.

Integration partners who move fast can win. Mo Shana'a, CEO of Morta, a Revizto integration partner, framed it well: "the model stops being a static deliverable and starts being something they use day-to-day." Nathan Beplate, Senior Digital Engineering Manager at John Holland Group, was more concrete, calling out that "direct, structured access to our own data removes a large portion of time-intensive manual interactions, effort we can no longer afford given the scale and pace of delivery." That's a customer telling you the labour arbitrage is already baked in.

The category most exposed is any AECO software vendor selling an "AI Copilot" as a premium SKU. If your differentiation was that you'd wired a chatbot to your own dataset, and now the underlying data platform is exposing the same surface to every LLM on the market, your moat just got a lot shallower. My take: expect a wave of "AI feature" repricing across the AECO sector by end of 2026.

Playbook for Data Teams

If you're a platform lead or head of data at an AECO organization, three things belong on this quarter's roadmap.

First, inventory every workflow that currently exports Revizto data to a spreadsheet or a BI extract. Each one is a candidate for replacement with a direct API call or an MCP-mediated prompt. The wins are largest where a human is currently the integration.

Second, get the OAuth 2.0 app registration into your identity provider now. Standards-based auth is only useful if it's plumbed into your SSO, your audit logs, and your offboarding process. The Developer Portal makes this a self-serve job for account admins, which is a gift, but only if someone owns it.

Third, decide where the semantic layer lives. If the LLM is going to query object properties, you need to agree on what "fire rating" means across projects and how quantities are normalized. Otherwise you get confident, fluent, wrong answers. This is the part where it all falls over for teams that skipped the boring work.

For CIOs specifically, the sovereignty story is the one to internalize. The architecture keeps sensitive project data inside the customer's control while letting the AI platform reach in for a specific query. That's the compliance framing you'll want when the risk committee asks why an LLM is touching structural drawings.

Key Takeaways

• Revizto's MCP Server, Model & Object Properties API and Developer Portal ship together under the existing subscription, removing the usual per-feature paywall friction.
• The Bridging the Gap Report 2026 numbers (32% capacity-constrained, 96% worried about data control, 39% simplifying their stack) are the market conditions this launch was built for.
• OAuth 2.0 replacing manual access codes is the quiet but important detail: it's what turns an integration story into an enterprise-buyable one.
• Middleware vendors and premium "AI Copilot" SKUs in AECO are the most exposed; integration partners who build workflow-specific tooling on top can win.
• The interchange-rail analogy holds: whoever owns the standard connection between project data and LLMs owns the next decade of AECO analytics, and Revizto has just laid track.