Problem
Agents currently lose all accumulated context when a session ends. This means the same constraints, preferences, and learned facts have to be re-established every time — creating repeated overhead and making the agent feel unreliable.
Goal
Memory written during a session should survive session boundaries and be available on the next run. The agent should be able to pick up where it left off without re-learning what it already knows.
Acceptance criteria
- Memory written in session N is retrievable in session N+1
- Retrieval respects recency and relevance, not just existence
- No manual export/import step required from the user
Problem
Agents currently lose all accumulated context when a session ends. This means the same constraints, preferences, and learned facts have to be re-established every time — creating repeated overhead and making the agent feel unreliable.
Goal
Memory written during a session should survive session boundaries and be available on the next run. The agent should be able to pick up where it left off without re-learning what it already knows.
Acceptance criteria