Skip to Content
Deploy & ConfigureOverview

Deploy & Configure

This section is for the DevOps, IT, and AI engineers in charge of setting up auxilia for a team (or for a solo developer running it on their own infrastructure). Everything a workspace admin needs to get auxilia running, connected to your data, and observable lives here.

If you only need a quick local run to try the product, head to Get Started. For how to build and configure agents once auxilia is up, see Build Agents.

What “deploying auxilia” actually involves

auxilia is composed of four services. You’ll decide where each one runs, how they talk to each other, and which external integrations are turned on.

ServiceTechnologyPurpose
BackendFastAPI + LangGraphAgent runtime, MCP client, auth, API
WebNext.js 16 + React 19User interface and backend proxy
DatabasePostgreSQL 17 + Redis 7Persistence, checkpoints, OAuth token storage
SandboxOpenSandbox (optional)Isolated code execution for agents

The repository ships with everything you need to build and run these:

backend/Dockerfile       # FastAPI backend
web/Dockerfile           # Next.js frontend
docker-compose.yml       # production compose file
docker-compose.dev.yml   # postgres + redis for local dev

The full setup checklist

A typical deployment goes through these steps — each one has its own page:

  1. Environment variables — pick a database, configure secrets, choose at least one LLM provider
  2. Deploy the services — Cloud Run or VM / Docker host
  3. User management — create the first admin, invite teammates, set up Google OAuth, mint personal access tokens
  4. MCP servers — register the tools (Notion, HubSpot, BigQuery, …) you want agents to use
  5. Sandbox — run OpenSandbox if you want agents to execute code
  6. Observability (Langfuse) — trace LLM calls and tool calls
  7. Slack integration — connect your Slack workspace

Steps 1–3 are mandatory. Steps 4–7 are additive — enable only what your team needs.

Migrations

Alembic migrations run automatically at backend startup (alembic upgrade head is invoked by the entrypoint). No extra step is needed on deploy — just restart the backend when pulling a new image.

Reverse proxy / HTTPS

The web container listens on port 3000 and proxies /api/backend/* to the backend on port 8000. Put a TLS-terminating reverse proxy in front (Cloud Run, Nginx, Caddy, Cloudflare, …) and set COOKIE_SECURE=true, FRONTEND_URL=https://your-domain.

Get StartedEnvironment Variables