@arcforge/engines

Provider connections for the Axon agent runtime.

An engine is a source of inference and nothing more. Axon's loop, context assembly, tool dispatch and policy enforcement are identical regardless of which provider answers — changing providers changes how an agent thinks, never how it works.

Supports Anthropic, OpenAI, Google, Groq, Cerebras, DeepSeek, Mistral, Moonshot, OpenRouter, Perplexity, xAI, Z.AI, Ollama for local models, and Axon's own managed inference.

npm install @arcforge/engines

Most people never install this directly. It arrives as a dependency of the Axon framework when you scaffold an agent with axon init.

Documentation

Full documentation is at axon.arclabs.it/docs/v2.

• Providers
• How a role becomes a model
• Engines

License

Proprietary. © ArcLabs