MESA - Memory Engine for Structured Agents
A durable, tenant- and dataset-isolated memory engine for autonomous AI agents, with a compatibility runtime that stores raw memories and an in-development release that adds source provenance, validation and combined SQL, vector and graph storage.
MESA is a memory engine built to give autonomous AI agents durable, tenant- and dataset-isolated storage. Its current compatibility runtime is a lexical core that commits accepted records as durable raw memories without calling a model or an external provider by default, running as separate API and worker processes over a persistent data volume.
A release candidate under development adds canonical source provenance, selectable validation, and combined SQL, vector and graph storage in a single storage-owner process, with model and external-provider access requiring explicit configuration rather than being on by default. The project states plainly that this newer runtime is not yet certified for production use. It can be run from a checked-out source tree or with Docker Compose, and exposes an HTTP API with interactive documentation.