Deploy and Host Mem0 on Railway

Mem0 is an open-source memory layer for AI agents and apps. It extracts durable facts from conversations, stores them as embeddings, and lets your agent recall them per user, agent, or session through a simple REST API. This template deploys the official self-hosted Mem0 server with its web dashboard and a pgvector Postgres database.

About Hosting Mem0

Hosting Mem0 means running the FastAPI REST server next to PostgreSQL with the pgvector extension, which holds both the memory embeddings and the app tables for users, API keys, request logs, and runtime settings. The dashboard is a separate Next.js service that talks to the API over its public HTTPS domain, and the API's CORS allow-list contains only the dashboard origin. Both app services are built directly from the upstream repository (server/ and server/dashboard/), because Mem0 does not publish a maintained prebuilt image for the current server. Authentication is on by default: the first visit to the dashboard runs a setup wizard that creates the admin account, and a generated ADMIN_API_KEY lets you call the API immediately.

Common Use Cases

Agent memory: give chatbots and autonomous agents long-term memory across sessions
Personalisation: remember user preferences, facts, and history per user_id
Multi-agent state: scope memories by agent_id and run_id
Semantic recall: POST /search returns the most relevant memories for a query
Auditing: the dashboard shows every API request, entity, and memory

Dependencies for Mem0 Hosting

PostgreSQL with pgvector (included, pgvector/pgvector:pg17)
OpenAI API key (required; the server initialises its OpenAI LLM and embedder at startup and will not boot without one)
Anthropic or Google keys (optional; switch providers later in the dashboard Configuration page)
No Neo4j and no Redis are needed for the REST server

Deployment Dependencies

Implementation Details

The API service installs the psycopg[binary] Postgres driver (upstream's Dockerfile currently omits it), runs alembic upgrade head so the app tables exist on first boot, then starts uvicorn on 0.0.0.0 so Railway's healthcheck can reach it. Memory history lives on a volume at /app/history. Both app services build from the upstream main branch, so a redeploy picks up upstream changes. Postgres holds the memories vector table and the app tables in one database.

PORT=8000
OPENAI_API_KEY=
JWT_SECRET=${{secret(64)}}
ADMIN_API_KEY=${{secret(32)}}
POSTGRES_HOST=${{Postgres.PGHOST}}
DASHBOARD_URL=https://${{Mem0 Dashboard.RAILWAY_PUBLIC_DOMAIN}}
NEXT_PUBLIC_API_URL=https://${{Mem0 API.RAILWAY_PUBLIC_DOMAIN}}
API_INTERNAL_URL=https://${{Mem0 API.RAILWAY_PUBLIC_DOMAIN}}

After deploy, open the dashboard URL and complete the setup wizard, then create a per-user API key under API Keys. Store a memory with POST /memories and the X-API-Key header, and read it back with GET /memories?user_id=.... Interactive API docs live at /docs on the API service.

Why Deploy Mem0 on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Mem0 on Railway, you get the API, dashboard, and pgvector database wired together with generated secrets, health checks, managed SSL, and private networking to the rest of your project.