Memfleet MCP server

What it does

memfleet provides shared, governed memory for fleets of AI agents running on MCP. Agent contributions pass through a judgment step before entering shared memory, with full provenance tracking and operator visibility. It is for teams running multiple AI agents that need to retain and share lessons across sessions.

• Multiple agents collaboratively solve problems by reading a composed perspective of shared memory relevant to their scope, without exhausting context windows.
• Operators monitor and steer agent memory directly from a web Console, seeing exactly what any agent can access and controlling what enters shared memory.
• Agent contributions are judged and recorded with provenance before becoming binding directives or informative context, ensuring memory quality and traceability.

Tools this server exposes

Tool	Does
read_perspective	Fetch the agent's composed view of shared memory scopes it can access.
list_scopes	List the scopes available in the workspace.
read_scope_summary	Get a summary of a specific scope's contents.
read_record	Fetch the provenance record for scope entries.
contribute	Submit a contribution to shared memory for judgment.
publish	Publish a contribution to shared memory after judgment approval.
withdraw	Remove a contribution from shared memory.
session_stats	Retrieve statistics about the current session.
session_closeout	Close out the current session.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.