Installation

AMFS is distributed as several packages so you can install only what you need. The fastest way to get started is with Docker — no Python required.

Get a running AMFS server in one command:

docker run -p 8080:8080 -v amfs-data:/data ghcr.io/raia-live/amfs

Or with Postgres (full-text + vector search):

docker compose up

See the Docker & Kubernetes guide for the full setup.

Python SDK

The core SDK includes the AgentMemory class and the filesystem adapter:

pip install amfs

Requires Python 3.11 or later.

Optional packages 

pip install amfs-adapter-postgres   # Postgres adapter (full-text + vector search)
pip install amfs-adapter-s3         # S3-compatible adapter (AWS, ACS, MinIO, R2)
pip install amfs-http-server        # HTTP/REST API server
pip install amfs-cli                # CLI tools
pip install amfs-mcp-server         # MCP server for AI coding agents

TypeScript SDK 

npm install @amfs/sdk

Framework Integrations

Install the integration package for your framework:

pip install amfs-crewai       # CrewAI
pip install amfs-langgraph    # LangGraph
pip install amfs-langchain    # LangChain
pip install amfs-autogen      # AutoGen

Initialize a Project

After installing, initialize AMFS in your project directory:

amfs init

This creates:

Path Purpose
amfs.yaml Configuration file
.amfs/ Local data directory (filesystem adapter storage)
.gitignore Updated to exclude .amfs/

You can skip amfs init if you’re using the Postgres adapter or passing configuration programmatically — the SDK works without a config file using sensible defaults.

Verify Installation 

from amfs import AgentMemory

mem = AgentMemory(agent_id="test")
entry = mem.write("test", "hello", "world")
print(entry.value)  # "world"
print("AMFS is working!")

Next Steps