Mireye is a location-infrastructure API for physical-world AI agents that need structured decisions about places in the United States, built by a Y Combinator Summer 2026 team. Instead of forcing a model to guess addresses and risk from unstructured web text, Mireye exposes 366 indexed fields through one API and an MCP server, so agents reason about the real world from verified data rather than memory.
Core Features
- A single REST API plus an MCP server, so both classic code and agent frameworks can call the same location data.
- 366 indexed location fields covering risk, property, and logistics signals, with the option to request custom fields.
- Insurance risk screening that scores a property before a policy is written.
- Proptech address management that normalizes and enriches messy location records.
- Robotics warehouse selection that ranks sites by physical and operational fit.
Use Cases
- Insurance underwriters automating property risk checks at scale.
- Real-estate and logistics teams cleaning and enriching address data.
- Robotics and fulfillment agents choosing warehouse locations from structured signals.
Pricing
Mireye is delivered through its API and MCP server on a usage-based plan; the company has not published flat per-call rates, so teams should expect volume pricing tied to field queries. Early access runs through the YC batch channel.
Our Take
Best for agent builders who need trustworthy U.S. location facts rather than model guesswork. The trade-off is scope: the 366-field coverage is U.S.-focused, so global physical-world agents will need a second data source. Compare other Productivity and agent tools.
