No Trust Falls: TestFit's MCP Keeps AI Contained in Reality

TestFit, the real estate feasibility platform used to evaluate more than 3,200 deals per week, today announced MCP (Model Context Protocol) support across its platform. Starting August 5, customers can connect the AI assistant of their choice, including Claude and ChatGPT, directly to their TestFit studies.

TestFit already runs a deal’s full workflow inside one engine: the developer's investment thesis, the architect's unit mix and massing, the civil engineer's grading and circulation, the general contractor's cost and constructability. MCP gives that engine a new way to direct it. Tell it to lay out a 300-unit, four-to five-story multifamily building that provides a 1.2 parking ratio, meets zoning requirements and hits a target yield. Nudge the unit mix. Compare two scenarios side by side and more. TestFit's engine still generates the layouts and the real quantity takeoffs, no hallucinating.

Built for Every Stakeholder in a Deal. Now You Just Tell It Yours.

Encoding the developer’s, architect’s, civil engineer’s, and GC’s logic into a single engine is what makes a TestFit answer defensible to a lender or a planning board, and what has made the tool itself a lot to learn.

"TestFit was built from nearly every piece of constructive feedback that the market wanted in its feasibility software. Often this meant hearing from brokers, developers, architects, cities, and even tenants themselves," said Clifton Harness, CEO and co-founder of TestFit. "TestFit has always carried the complexity of a deal so no single person has to. MCP enables how you direct it. A developer can tell it to hit a yield target. A civil engineer can tell it to solve the grading. Everyone gets the outcome their job needs, as fast as they can ask for it."

You don't need to know all 20-some disciplines, or hunt for the specific tool built for yours. Tell TestFit what your job needs, and it delivers it.

“We didn't build TestFit on AI. We built the thing AI needed to run on," said Laura Paciano, Chief Growth Officer at TestFit. "MCP skips the learning curve entirely. If you can say what you need out loud, you can use TestFit."

The AI Pushes the Buttons. The Engine Still Does the Work.

A new wave of AI assistants is showing up inside design and engineering software. Ask one to check an ADA clearance or sketch a floor plan, and a language model reads the file and generates the answer itself.  That's a trust fall: the model doing the reasoning is the same model you're trusting not to invent an answer that merely sounds right. 

TestFit's MCP separates those jobs. The AI assistant doesn’t generate the geometry or the numbers behind it; it just tells TestFit’s deterministic engine what outcome to hit, the same engine that has always powered the platform.

"Winning in real estate means building the best possible idea for this moment in time," said Harness. "TestFit is a due diligence and compliance machine, solving over 900 problems at once for the benefit of no less than 12 professions."

Users used to flip every one of those toggles by hand. Now the AI does it for them, and TestFit's engine still creates what's feasible while the human stays in command of what’s real.

Built on the Open Standard

MCP is the open protocol for connecting AI assistants to external tools and data, and it's been adopted broadly across the AI industry. TestFit built on that open standard so customers connect the assistant they already trust, Claude, ChatGPT, or others, with no proprietary lock-in. Because that connection runs through the customer’s own AI account, the conversation stays there, not on TestFit's servers and not trained on.

Rollout begins August 5, 2026.