Memanto
An agent that manages the shared memory of a fleet of AI agents: extracting durable facts from their conversations, merging duplicates, resolving contradictions, forgetting stale information, and briefing each agent before it acts.
Memanto is a memory agent that runs alongside a fleet of other AI agents rather than acting as a library they call directly. It watches their interaction streams and extracts durable knowledge, such as decisions, preferences, facts and failures, instead of archiving full transcripts.
It consolidates duplicate or fragmentary memories into one canonical estate, and when new information contradicts old information it supersedes rather than overwrites, so what is believed now and what was believed earlier remain distinguishable. Decay, expiry and deliberate deletion are handled as policies rather than manual cleanup, and before an agent acts, Memanto hands it the minimal relevant slice of the shared memory instead of requiring the agent to query it directly.
It installs with a pip package and is commonly paired with agent frameworks such as CrewAI and LangChain. It is released under the MIT license.