rewind
A personal data API that unifies my music, movies, books, runs, collections, and the concerts and games I go to, pulling from ten external services into one schema.
Claude on iOS querying the Rewind MCP server for attended games, recent watches, and top albums. Three domains, one schema.
Rewind is a personal data API. It pulls from ten external services and normalizes everything I listen to, watch, run, read, and collect into a unified schema, then serves the result through a REST API and an MCP server Claude can query directly.
Aggregating that data was actually relatively straightforward, and was the thing that unlocked the richer cross-domain queries once everything lived in a shared schema. Concerts and sporting events get pulled in automatically from Calendar and Gmail, then cross-referenced with the artists I already listen to and the players whose stats I follow. I can ask Claude how a Mariners hitter has performed in the games I've actually been in the stands for, or which artists I've seen live but barely played at home. None of that is a feature any single platform offers, because none of them have the other half of the data.
It's built on Hono running on Cloudflare Workers, with D1 for storage, R2 for images, and Drizzle for type-safe queries. End-to-end type inference via Hono RPC means consuming apps get a fully typed client with zero codegen. The whole API is documented at docs.rewind.rest with the same shape a production service would have, including versioned endpoints and an interactive OpenAPI spec.