Coral is an open-source, local-first SQL data layer that lets AI agents query APIs, databases, and files as read-only tables using standard SQL while handling auth, pagination, rate limits, schema mapping, caching, and cross-source JOINs so you don't need custom connectors or ETL; it plugs into agents via the CLI or MCP. By surfacing cached, semantically enriched context, schema hints, and query history to agents, Coral improves agent accuracy and cost-efficiency, enforces scoped permissions and governance, and keeps data inside your infrastructure—making it ideal for teams building production AI agents that must access heterogeneous live sources reliably.