Deploy Honcho Memory Server
Agent memory API with deriver worker and pgvector, JWT auth on by default
Postgres
Just deployed
Honcho Deriver
Just deployed
Just deployed
Deploy and Host Honcho Memory Server on Railway
Honcho is an open source memory layer for AI agents by Plastic Labs. Your app stores workspaces, peers, sessions and messages through a REST API, and a background worker reasons over them to build long term memories, peer cards and summaries that agents can query in natural language.
About Hosting Honcho
Honcho runs as an API server and a background worker (the deriver) that share one Postgres database with pgvector. This template uses the official ghcr.io/plastic-labs/honcho:v3.6.1 image for both, runs database migrations on boot, and keeps Postgres on Railway's private network. JWT auth is on from the first deploy with a generated signing secret, so the public API never accepts anonymous writes. Redis is optional upstream (cache only) and is left out.
Common Use Cases
- Agent memory: give chatbots and agents durable memory of each user across sessions
- User modeling: build peer cards and representations of users and agents from conversations
- Context retrieval: fetch session context and summaries sized for your model's window
- Dialectic queries: ask Honcho what it knows about a user in plain language
- Self hosted alternative to Honcho Cloud: keep conversation data in your own database
Dependencies for Honcho Hosting
- PostgreSQL with pgvector: included (
pgvector/pgvector:pg17with a volume) - LLM API key: optional at boot. Needed for memory reasoning, summaries, dreams, dialectic chat and embeddings (
LLM_OPENAI_API_KEY, or other providers via per module settings)
Deployment Dependencies
Implementation Details
PORT=8000
AUTH_USE_AUTH=true
AUTH_JWT_SECRET=${{secret(64,"abcdef0123456789")}}
DB_CONNECTION_URI=postgresql+psycopg://${{Postgres.POSTGRES_USER}}:${{Postgres.POSTGRES_PASSWORD}}@${{Postgres.RAILWAY_PRIVATE_DOMAIN}}:5432/${{Postgres.POSTGRES_DB}}
LLM_OPENAI_API_KEY= # optional
First steps after deploy:
- Open a shell with
railway ssh --service Honcho, then runcd /app && python scripts/generate_jwt.py --admin --print-onlyto get an admin key. - Call
POST /v3/keys?workspace_id=with the admin key to mint a key scoped to one workspace for your app. - Point the Honcho Python or TypeScript SDK at your Railway domain with that key.
- Add
LLM_OPENAI_API_KEYon the Honcho service and redeploy both Honcho services to turn on memory reasoning and message embeddings.
Without an LLM key, Honcho still stores workspaces, peers, sessions and messages; message embeddings stay off until a key is present. Rotating AUTH_JWT_SECRET invalidates every issued key.
Why Deploy Honcho 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 Honcho on Railway, you get the API, the deriver worker and pgvector Postgres wired together with generated secrets, auth on by default, health checks, managed SSL and private networking to your agents.
Template Content
Postgres
pgvector/pgvector:pg17Honcho Deriver
ghcr.io/plastic-labs/honcho:v3.6.1
