Memanto

Memanto is a memory agent: a companion agent that runs beside your other coding agents and manages their shared long-term memory. Instead of only storing what agents write, it decides what to keep, merges duplicates, resolves contradictions between agents, expires stale memories by policy, and briefs each agent with the relevant slice of memory before it acts.

It observes agent conversations and extracts decisions, preferences, facts, and failures rather than archiving whole transcripts. When new knowledge contradicts old, it supersedes the old memory while keeping a record of what was believed and when, so recall can answer both what is true now and what was believed at a past date or before a release. A scheduled daily run curates new memories, merges duplicates across agents, and flags contradictions for review.

Memories are typed into 13 categories, such as instruction, fact, decision, goal, and preference, and each carries a confidence score, source, provenance, and timestamp. Each agent has its own namespace. A memory is active or expired, expiry comes only from a retention policy you set, and expired memories stay recallable with a label and can be restored.

Recall runs on a bundled semantic engine, either as a local Docker container with Ollama models and no account or as a free cloud service, and memories are searchable as soon as they are written. The whole memory estate exports to the Open Knowledge Format as plain Markdown, and Memanto imports from other memory systems. Python, TypeScript, and Go SDKs, a REST API, a CLI, and a local web dashboard are available.