InfrOS launches MCP Server for AI-assisted cloud infrastructure

InfrOS is rolling out a Model Context Protocol server in preview for eligible customers, letting engineering teams design, compare, and validate cloud infrastructure inside supported AI assistants and development tools. The launch aims to reduce guesswork before deployment by combining architecture design, Terraform generation, and proof-before-deploy checks.

Why it matters: - Engineering teams can move from cloud design to validated infrastructure-as-code inside AI assistants, instead of bouncing between separate tools. - The preview is aimed at reducing deployment risk by checking cost, performance, connectivity, and provisioning before production rollout. - Teams keep control of repository changes and production deployment, which limits what the AI workflow can do on its own.

What happened: - InfrOS announced the InfrOS Model Context Protocol Server on November 9, 2026, in Tel Aviv, Israel. - The MCP Server brings InfrOS architecture design, infrastructure evolution, and validated infrastructure-as-code workflows into supported AI assistants and development environments. - The launch is in preview for eligible customers. - Design runs and Terraform generation consume organization credits.

The details: - The MCP Server lets teams create design specifications, update requirements, and identify missing information before a design run. - The tools can prompt the assistant to ask for user decisions instead of guessing when requirements are incomplete. - Users can design around an existing organizational landing zone or start from scratch. - The system can compare a landing-zone-based design with a greenfield alternative. - The platform can return architecture proposals with projected monthly costs in U.S. dollars, quality ratings, advantages, disadvantages, and a recommendation. - Users can inspect individual components, including services, configurations, regions, and costs. - The MCP Server can generate and retrieve Terraform for supported proposals. - The assistant can place those files in the user’s workspace using its own authorized file-access capabilities. - Authorized organization administrators can manage landing zones and standing constraints. - Where GitHub integration is enabled, administrators can run read-only repository analyses. - Those analyses can surface Terraform versions, providers, environment structure, and secrets-handling patterns InfrOS has learned. - InfrOS says supported AI clients, setup instructions, and access requirements are available here.

Between the lines: - InfrOS is positioning the MCP Server as a way to make AI-assisted infrastructure work more grounded and less speculative. - The emphasis on missing requirements, cost trade-offs, and recommendation logic suggests the product is built to steer users toward decision-making before code generation. - The governance model is restrictive by design: InfrOS can generate files and validate them, but it does not take over repo changes or deployment. - OAuth separates reading, editing, credit-spending, and organization administration, which keeps access tied to user permissions and enabled features. - Users can review connected assistants and revoke access from their InfrOS profile. - Tool-call records capture who acted, which client and tool were used, and the outcome, but not the full request and response contents.

What's next: - Eligible customers can start using the MCP Server in preview. - Teams that enable the workflow can use Prove Before Deploy to validate generated Terraform in InfrOS’s sandbox before production use. - Some setup and confirmation steps still require the InfrOS web application.

The bottom line: - InfrOS is trying to make AI-assisted infrastructure design more reliable by pairing generation with validation, benchmarking, and human-controlled deployment.