Skip to content

Laptop or dev box

The local path: run an open model on your own machine (no GPU needed) and drive Claude Code or an OpenAI client against it. Great for development, evals, and offline work.

Terminal window
atlas up --model qwen3-0.6b
# in another shell:
ANTHROPIC_BASE_URL=http://localhost:8080 ANTHROPIC_API_KEY=<printed key> claude

atlas up provisions the llama.cpp runtime on first use and serves on http://localhost:8080. The default API key is printed on first start.

One-shot, no daemon:

Terminal window
atlas run qwen3-0.6b "your prompt"

Prefer containers? The slim image does the same.

See the quickstart for the full first-run walkthrough.