MFS
A context harness for AI agents that presents code, memory, docs, chats and databases from many sources as one searchable, browsable file-like tree.
MFS (Multi-source File-like Search) gathers an AI agent's scattered context, such as repositories, past-session memory, SKILL.md skills, documents, chat and mail services, issue trackers and databases, under one shell where every source becomes a file-like tree under a stable URI. Two skills the agent loads bundle its commands: mfs-ingest registers and re-indexes connectors, and mfs-find searches and browses across what has been ingested, with search and grep for fast lookup and ls, cat, tree, head and tail for precise navigation.
It runs as a server, installable from PyPI with a uv tool or via Docker Compose, with a separate CLI client installed through Cargo. A roughly 600 MB embedding model downloads on first use and is cached for subsequent runs.
It is written in Python and licensed under Apache-2.0.