Docker

run engram as a container with persistent data.

quick start

git clone https://github.com/raya-ac/engram.git
cd engram
docker compose up -d

web dashboard at http://localhost:8420.

docker compose

services:
  engram:
    build: .
    ports:
      - "8420:8420"
    volumes:
      - engram-data:/data
      - ./config.yaml:/app/config.yaml:ro
    environment:
      - ENGRAM_DB_PATH=/data/memory.db
      # - VOYAGE_API_KEY=your-key
      # - OPENAI_API_KEY=your-key
    restart: unless-stopped

volumes:
  engram-data:

configuration

mount your config.yaml to /app/config.yaml. set API keys via environment variables.

the database lives in the engram-data volume at /data/memory.db.

auth

lock down the web dashboard:

# config.yaml
web:
  auth_token: "your-secret-token"

access with ?token=your-secret-token in the URL or Authorization: Bearer your-secret-token header.

MCP over SSE

for remote MCP access (not stdio):

services:
  engram-mcp:
    build: .
    ports:
      - "8421:8421"
    command: ["python", "-m", "engram", "serve", "--mcp-sse", "--port", "8421"]
    volumes:
      - engram-data:/data
    environment:
      - ENGRAM_DB_PATH=/data/memory.db

endpoints:

• POST /mcp — JSON-RPC
• GET /sse — SSE stream
• GET /health — health check

building the image

docker build -t engram .
docker run -p 8420:8420 -v engram-data:/data engram

the Dockerfile uses python:3.12-slim, installs all deps including API backends, and exposes port 8420.