Integration

Axocoatl × OpenRouter.

Every model on OpenRouter, available to every agent in Axocoatl. One API key, one config block. Per-agent model pinning, fallback chains, and the rest of the runtime — unchanged.

Why pair them.

OpenRouter routes your request to the right LLM provider: OpenAI, Anthropic, Mistral, DeepSeek, Llama hosts, dozens more. One API key, one URL, every model. You pay per token, you switch models without touching billing, and you can A/B two models on the same workflow without two integrations.

Axocoatl handles the orchestration. OpenRouter handles the inference. Pair them and you get a self-supervising runtime on your hardware that can ask any frontier model when it needs to, and a local model when it doesn't.

Set it up in 30 seconds.

Add an openrouter block to your axocoatl.yaml:

providers:
  openrouter:
    api_key: "${OPENROUTER_API_KEY}"

agents:
  - id: researcher
    provider: openrouter
    model: "anthropic/claude-3.5-sonnet"
  - id: summarizer
    provider: openrouter
    model: "openai/gpt-4o-mini"      # Mixed-model workflow, one config.

Put OPENROUTER_API_KEY=… in your .env. Restart the daemon. That's the whole integration.

Or do it through onboard.

The interactive setup wizard picks OpenRouter for you when you select "cloud, model of my choice":

axocoatl onboard

Patterns

Three setups we recommend.

Cost-tier routing

Use openai/gpt-4o-mini on every agent by default. Promote individual agents to anthropic/claude-3.5-sonnet when they need heavy reasoning. The lattice picks the right one per task.

Local + cloud hybrid

Ollama for cheap, repetitive agents (summarization, classification). OpenRouter for the one agent that needs frontier intelligence. Same workflow, two providers.

Failover chain

Primary provider = OpenRouter. Fallback = direct Anthropic credentials. If OpenRouter degrades, the daemon switches without dropping the run.

About OpenRouter's leaderboard.

OpenRouter publishes an app directory ranking the tools that route the most traffic through it. Axocoatl is listed there. If you set up your daemon with OpenRouter, your usage shows up under the Axocoatl entry — you don't have to do anything special.

We're not optimizing for that leaderboard at the expense of the local-first commitment. Ollama remains the default recommendation in axocoatl onboard. OpenRouter is the cloud option for teams who don't want to manage their own LLM keys per provider.

Get an API key. Set the config. Done.

OpenRouter's signup is free. The Axocoatl daemon is one curl-pipe-sh away.

OpenRouter key → Install Axocoatl