Memfleet

memfleet

Shared, governed memory for fleets of AI agents. Contributions pass judgment; every entry carries provenance; operators see everything.

This repository is the server home for the memfleet MCP server — com.memfleet/memfleet on the official MCP Registry. The platform lives at memfleet.com; the command-line client ships as memfleet on PyPI.

What memfleet is

memfleet is the hosted platform for Strata, the open-source memory engine. You create a Workspace (a full memory fleet), register agents that receive keys, and operate everything from a web Console while your agents read and contribute over MCP or REST. Judgment runs as a platform service.

Agent teams forget: every session starts from zero and lessons learned by one agent never reach the rest. With memfleet, agents send contributions; nothing writes to shared memory directly. A judgment step decides what enters and whether it lands as a binding directive or informative context. Memory is organized in scopes arranged in strata, each agent reads a composed perspective of what its scope can see, and the record keeps provenance for every entry. Operators manage the whole lifecycle from a web Console, including seeing any agent's exact perspective.

The MCP server

• Endpoint: https://memfleet.com/mcp (streamable HTTP)
• Auth: Authorization: Bearer — minted for a Registered Agent in your Workspace's Console; resolve it from an environment variable, never commit it.
• Tools: read perspective, list scopes, read scope summary and record, contribute, publish, withdraw, session stats, session closeout.
• REST: the same operations at /agent/v1 for anything that speaks HTTP.

Agents read and contribute over MCP or REST, so any MCP client can join the fleet; first-class setup ships for Claude Code today.

The server.json in this repository is the registry listing manifest for com.memfleet/memfleet.

Quickstart

pipx install memfleet

memfleet connect # once per machine — one browser approval memfleet run --profile 'Fleet/backend-bot' claude # per project / per session

connect links the machine to your account. run binds the session to a Registered Agent — creating it if it does not exist yet — writes the project wiring, and launches the command. The setup it writes is secret-free: no key is ever written to disk in plaintext.

What the fleet gets

• Governed memory, not a shared scratchpad. Every contribution passes judgment before it enters shared memory; every entry lives in a scope and carries provenance in the record.
• Memory that doesn't eat your context window. An agent reads its perspective: a compact, composed summary of what its scope can see, not a transcript dump. That keeps the tokens your agent spends on memory small, and reading a perspective is never metered.
• The operator is sovereign. See exactly what any scope's agents see, and steer memory directly. No hidden state.
• The engine is open source. Run Strata yourself, or let memfleet run the fleet for you.

Links

• Platform and Console: memfleet.com
• Pricing: memfleet.com/pricing
• Client on PyPI: pypi.org/project/memfleet
• Engine, open source: github.com/oren198/Strata

License

MIT