Start local and free with Ollama; switch to a cloud provider for production. The agent code stays the same.
| Provider | Best for | Privacy |
|---|
Step 4 uses the axiom engine, but it does not depend on the rest of Æon. Install aeon-axioms on its own and wrap the executor of any framework, LangGraph, CrewAI, an MCP host or an A2A endpoint, to gate tool actions deterministically, alongside the content filters and guardrails you already use.
Begin with Ollama for development, it runs entirely on your machine, so your data and your chaos experiments never leave it.