Auxot
Docs Pricing Login

Prerequisites

Auxot requires two backing services:

  • PostgreSQL 15+ — stores users, providers, agents, conversations, and configuration.
  • Redis 7+ — handles pub/sub for streaming responses and ephemeral session state.

The quickest path is Docker Compose, which handles both automatically. Otherwise, point Auxot at existing Postgres and Redis instances via environment variables.

Install Auxot

Quick Start (One-Liner)

curl -fsSL https://get.auxot.com/install | sh

This downloads the auxot-server binary, installs it, and starts Auxot with Docker Compose (Postgres + Redis included).

Manual (Inspect First)

If you prefer to inspect the compose file before running:

curl -fsSLO https://get.auxot.com/docker-compose.yml
docker compose up

See Deployment for production hardening and alternative deployment methods.

Docker Image (Direct)

docker pull ghcr.io/auxothq/auxot-server:latest

Start the Server

If running the binary directly:

auxot-server

On first launch, Auxot will:

  1. Run database migrations against the configured Postgres instance.
  2. Seed the admin agent — a built-in AI agent that helps you manage the system.
  3. Start listening on port 8080 (configurable via AUXOT_BIND_ADDR).

DATABASE_URL, REDIS_URL, and AUXOT_SECRET_KEY are required before startup.

First-Run Setup

Open your browser to http://localhost:8080. You’ll be greeted by the admin agent, which walks you through:

  1. Creating your admin account — sets your email and password.
  2. Installing a license key (optional) — paste a key from auxot.com to unlock paid features, or skip to use the free tier.
  3. Basic configuration — set your organization name and default routing preferences.

Add Your First Provider

The fastest way to start generating responses is to add a CLI worker running Claude Code. On any machine with the claude CLI installed:

# Install Claude Code on the server host
npm install -g @anthropic-ai/claude-code

# Authenticate (interactive)
claude login

CLI workers are configured through the admin agent or the web UI, then run as separate worker processes/containers that connect back to Auxot over WebSocket. Once Claude Code is installed and authenticated on the worker host, start one or more CLI workers and confirm they show as online.

Send Your First Message

Use the built-in chat UI at http://localhost:8080, or hit the API directly:

curl http://localhost:8080/api/openai/chat/completions \
  -H "Authorization: Bearer <your-api-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [
      {"role": "user", "content": "Hello! What can you help me with?"}
    ]
  }'

Setting model to "auto" lets the router pick the best available provider. You’ll see a streamed response routed through whichever provider is online — GPU first, then CLI, then cloud.

Next Steps