On July 28, Revizto announced three things that sound like plumbing and amount to a strategic position: an MCP Server, a Model & Object Properties API, and a Developer Portal. Together they let an AECO team point ChatGPT, Claude or Microsoft Copilot at live Revizto project data without building a custom integration first.
Every other major platform in this sector spent 2026 building agents that live inside its own product. Revizto has done close to the opposite. It has made its data legible to whichever model the customer already pays for.
What Actually Shipped
The MCP Server is the centrepiece. Model Context Protocol is the open standard for connecting language models to external systems, and Revizto’s implementation provides what the company describes as a secure connection between Revizto and any large language model, translating project data into a form a model can query — without a bespoke integration per tool.
The Model & Object Properties API exposes real-time model metadata: materials, systems, fire ratings, quantities. Revizto positions it for quantity takeoffs, dashboards, and feeding procurement systems. This is the part that turns a coordination tool into a data source other software can build on.
The Developer Portal is the least glamorous and possibly the most consequential. It replaces manually issued access codes with OAuth 2.0 app registration, so integrations become self-serve with standard administrative controls. Manual key distribution is exactly the friction that keeps enterprise IT from approving integrations at all.
All three are included under existing Revizto subscriptions rather than sold as an AI add-on — a pointed contrast with the metered-credit model the incumbents are moving toward.
“The organizations that outperform will be those that can securely connect intelligence to trusted, live project data.” — Marc Schütz, Chief Product Officer, Revizto
The Strategic Bet
There are two coherent answers to “how does AI reach construction data,” and the industry is now visibly split between them.
The first is the walled garden. Procore’s Digital Coworker packages put twenty-odd agents inside Procore, metered by credits, administered through a Procore console — and the company has just spent $845 million on DroneDeploy to feed them. Autodesk has done the same inside Forma. The argument is quality: a vendor that controls the model, the data and the interface can guarantee the answer, and can sell the outcome.
The second is the open pipe. Revizto’s argument is that the customer has already chosen its AI vendor — it is whatever the enterprise licence covers, usually Copilot or ChatGPT — and the platform’s job is to make its data reachable from there. The argument is leverage: you do not have to win the model war to be valuable in it.
The second bet is unusual in vertical software because it deliberately gives up the agent layer, which is where everyone assumes the margin will be. Revizto is betting that being the trusted source of live model data is a better long-term position than being one more chat interface. It is a plausible read of a market where, as McKinsey’s own sequencing suggests, the durable advantage accrues to whoever controls proprietary project data rather than whoever ships the slickest assistant.
The Names on the Release
Revizto brought a partner and a customer, which is more than most launches of this kind manage.
Morta, whose CEO and co-founder Mo Shana’a is quoted, is the integration partner — his framing is that Revizto is opening the data and Morta lets teams work with it inside processes they already run. On the customer side, Nathan Beplate, Senior Digital Engineering Manager at John Holland Group, the Australian contractor, says direct structured access to their own data “removes a large portion of time-intensive manual interactions.”
That last phrase is worth sitting with, because it is the actual use case and it is duller than the marketing. Nobody in the release claims the AI designs anything. The claim is that people currently spend hours manually extracting data they already own, and that this stops.
Revizto’s release also lists Jacobs, AECOM, AtkinsRéalis, Skanska and Stantec among its clients. Those are the company’s own customer references rather than participants in this announcement, and should be read as such.
What Could Go Wrong
The obvious risk is governance. The pitch is “connect your enterprise AI to live project data,” and the thing that makes an enterprise nervous about exactly that is where the data goes once a general-purpose model can read it. OAuth registration and admin controls are the right answer structurally, but the survey data suggests it is a live objection: 20% of construction respondents in DEWALT’s global study named privacy, cybersecurity and IP risk as a barrier to AI adoption. Revizto is asking those firms to open a pipe. Some IT departments will say no.
The second risk is that the open pipe is easy to copy. If exposing an MCP server becomes table stakes, the differentiation collapses back to the underlying data and the coordination product that generates it — which is fine for Revizto, but means the announcement is a fast-follow target rather than a moat.
What to Watch
Whether the incumbents match it. An MCP server on Procore or Autodesk data would be a genuine concession that customers will not use one vendor’s agent for everything — and the moment that happens, the walled-garden thesis is in trouble. Watch also whether anyone reports a real workflow running through this, rather than a quote about removing manual interactions. The gap between “the API exists” and “the estimator uses it on Tuesday” is where most interoperability announcements go to die.
