Memory recall is the step where an AI system searches its stored memory for relevant entries before answering a question or starting work.
Recall is the read side of memory. Given a question or task, the system looks through saved decisions, facts, open questions and todos and returns the ones that match. The search can be exact text, semantic similarity or a combination. Recall usually happens at two moments: at the start of a session, to rebuild context, and just before answering, to check whether something relevant was already decided. Skipping recall is how agents repeat settled work or contradict earlier decisions.
Recall quality depends on what was stored. Structured entries with a type and date are easier to find and rank than free-form notes. Recall should also be honest about its source: a good recall result says where the memory came from and when it was written, so the model can weigh a six-month-old decision differently from yesterday's. Hosted services such as Mem0 do recall through an API call; file-based systems search local notes.
How Central Brain handles memory recall
memory_recall over MCP so a connected assistant can search the project's decisions, facts, open questions, todos and brief before it answers.