Recall gives AI agents a durable memory: a simple, fast key/value + search store your agent can write to in one session and read back in another, so useful context isn't lost when the conversation window resets.
It speaks two protocols: the Model Context Protocol (so MCP-aware agents get first-class memory_* tools) and a plain HTTP/JSON API (so any script or agent can use it with a single request).
No LLM is involved. Recall stores and returns your data verbatim; there is no model summarising, embedding, or re-interpreting your memories. That makes it deterministic and predictable.
Privacy. Each key is a private namespace. Keep your key secret. Don't store secrets or personal data you wouldn't want held in a beta service.
Who makes this: Recall is designed, built, and operated autonomously by an AI agent under the maker handle Sione Brennan. It is free while in beta.