OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

MFS — Multi-source File-like Search

MFS is an open-source, self-hosted context retrieval tool for AI agents. Its server and CLI bring registered sources—including code, memory, skills, documents, messages and databases—into a unified file-like namespace that agents can search and browse.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

MFS — Multi-source File-like Search — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use MFS when an agent must retrieve relevant material from context spread across multiple connected sources, or when you want to build an agent workflow around a shared searchable namespace.

Who it is for

  • Developers building or operating AI agents that need context from multiple sources.
  • Teams seeking a unified search and browsing interface for code, project documents, conversation history or database records.

Documented uses

  • Register and index context sources, then re-sync them as needed.
  • Search one source or fan a query across registered sources.
  • Browse search results and read files or full database records through shared file-like locators.
  • Give an agent skills for source ingestion and context retrieval.
Before you choose

Requirements & limits

MFS is self-hosted and open source under the Apache-2.0 license. The documented setup uses a server and CLI; an agent-skill installation is also provided.

  • The service is run as a server, with a separate CLI client and optional agent skills.
  • The README documents uv-based or Docker Compose server setup and Cargo-based CLI installation.
  • The first add operation downloads an approximately 600 MB embedding model, according to the README.
  • For Docker, the sample local directory is uploaded with mfs add --upload; the uv option reads local paths directly.
  • The README names broad source types including repositories, memory files, PDFs, chat, issue trackers, databases and object stores.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up MFS and search a local sample tree

Run the MFS server and CLI, index a small local folder, then search it.

Before you start

  • Python 3.10 or later for the documented server environment.
  • uv for the tool-install option, or Docker Compose for the container option.
  • Cargo to install the MFS CLI.
  • A local folder to index; the README uses /tmp/hello-mfs.

Steps

  1. Start the server with the command below, or clone the repository and start its Docker Compose configuration.

    Documented command
    uv tool install mfs-server && mfs-server run

  2. In a separate terminal, install the client with the command below; set MFS_API_URL to the server address and provide MFS_API_TOKEN when required by the chosen server option.

    Documented command
    cargo install mfs-cli

  3. Create a sample directory and text file, then run mfs add on the directory. For the Docker option, use --upload to send the tree to the container.

  4. Run mfs search with a natural-language query and the indexed path.

Expected outcome

A running MFS service with the sample tree indexed and searchable through the CLI.

Guide limitations

  • The first Docker build takes a few minutes; the embedding model downloads on the first mfs add and is cached in the mfs-data volume.
  • The README documents an additional macOS quarantine-removal command if the CLI is flagged as from an unidentified developer.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
12 October 2026
Original website title
zilliztech/mfs · Official homepage title · retained in its source language
How we review sourcesSuggest a correction