Pirros has introduced a Model Context Protocol connection that gives AI assistants direct access to architectural and engineering details and Revit families stored across a firm’s Pirros workspaces. The Pirros MCP is designed to make institutional design knowledge accessible through AI tools that employees already use, including ChatGPT, Claude, Cursor, and Codex.
Users can ask an assistant in plain English to locate typical details, retrieve drawing PDFs, search for Revit families, identify dependencies between families, review revision histories, and create collections called stashes. For example, a designer can check whether a particular door family already exists before creating another version. The system can search by attributes such as name, category, and Revit version while identifying possible duplicates.
The MCP differs from tools that connect AI directly to an open Revit model. Autodesk’s Revit Public MCP Server, for example, provides read-only access to the model running on a user’s computer. Pirros MCP instead searches design knowledge accumulated across previous projects and stored within Pirros. For work inside an active Revit model, Pirros offers Mira, its AI agent, and Pirros in Revit.
Administrative controls determine what information AI assistants can reach. Firms can create credentials for individual AI tools, specify accessible workspaces, and control read, write, and download permissions. Credentials can also be narrowed to selected content. Pirros says company data is not used to train AI models and notes that its platform is SOC 2 Type II compliant.
The MCP forms part of Pirros Connections, which also includes an API, shared workspaces, and SAML single sign-on with role-based access. Together, these capabilities aim to reduce the effort required to find and reuse existing design knowledge while keeping access under firm-level control.
