Relay on your
infra, full
control.

When compliance requires data to stay inside your perimeter, when you need your own contract with Anthropic directly, when you want full logs in your own SIEM — we deploy Agent Nine on your side. Docker-compose with Postgres and the admin panel. Keys and data are yours alone; we support it at the code level.

The stack that runs on your infra.

Delivery is a docker-compose with four services plus our client app, pointed at your domain instead of agentnine.io. No proprietary "enterprise builds" — the same code as our production, just under your control.

01 · Relay

Fastify + WebSocket server

A Node app in Docker, listening on the port you choose. Handles auth, the LLM proxy, MCP, quotas, sessions. Source is opened under the contract.

02 · Postgres

Your Postgres 16

Stores users, sessions, transcripts, api_calls. Connect your existing cluster or run it as part of the compose stack. Backups follow your processes.

03 · LLM routing

Your providers, your keys

The relay routes LLM requests by config: list your providers (Anthropic direct, OpenAI-compatible and any compatible ones), order and fallback. Keys are environment variables, never in code.

04 · Admin panel

User management

A web interface to invite, block, view spend and set per-user quotas. Installed alongside the relay, access via your SSO.

05 · Desktop client

Built for your relay

A custom MSI/.deb with your domain in tauri.conf.json and your updater channel. Auto-update pulls from your server, no trips to agentnine.io.

06 · Support

No "not our problem"

We support the product at the code level: bug fixes, security patches, minor releases. You own the infra. Boundaries are written into the SLA.

Who owns what.

Our side

What we provide

  • Code: relay, desktop client, admin panel
  • Updates and security patches (LTS channel of your choice)
  • Team onboarding and documentation for your stack
  • Support over Slack / email / video call — per contract
  • Recommendations for production infra (HA, backups, monitoring)
  • Help investigating incidents when they involve our code
Your side

What stays with you

  • Hardware and hosting (bare metal or cloud — your choice)
  • DNS, TLS certificates, firewall
  • Postgres backups per your processes
  • SSO/IdP integration (Keycloak, Authentik, your OpenID)
  • LLM provider API keys and paying them directly
  • Monitoring (we expose Prometheus / Grafana metrics)

Let's scope the deployment.

Tell us your team size, stack and compliance requirements. We’ll send a scope with an effort estimate (usually 1–3 weeks), a dependency list, a draft contract and a machine spec for your load profile.