add-atomic-chat-tool

You connect NanoClaw to local Atomic Chat models by integrating a stdio-based MCP server that exposes tools for model discovery and chat completion generation via the app's OpenAI-compatible API. This allows on-device inference without external API calls.

Yes, you can use local AI models with NanoClaw without external cloud APIs by connecting the agent to the Atomic Chat desktop application. This setup satisfies functional requirements for private, on-device query processing like summarization and translation.

The best way to configure an OpenAI-compatible API for a containerized agent is using a stdio-based MCP server integration. This approach supports configurable host and authentication overrides to securely access local models.

Yes, NanoClaw supports discovering local models hosted by the Atomic Chat desktop app through an MCP server tool. It lists available local models and sends generation requests to them using the app's OpenAI-compatible API.

You set up local model access by applying additive, idempotent code edits to your existing NanoClaw configuration. This seamless integration adds the MCP server without requiring branch merges, enabling immediate access to Atomic Chat models.

Atomic Chat is limited to the local models available on your device and requires the desktop app to be running. It is designed for academic and personal workflows needing private summarization, translation, and general query processing.