LM Studio Claude API setup guide

The core issue looks less like missing compatibility and more like scattered docs plus performance-tuning friction.

• –LM Studio now documents both OpenAI-compatible and Anthropic-compatible endpoints, so Claude-like clients can usually just point `ANTHROPIC_BASE_URL` at `http://localhost:1234`.
• –The Claude Code walkthrough is the clearest start here path, with `ANTHROPIC_AUTH_TOKEN=lmstudio` and `claude --model openai/gpt-oss-20b`.
• –The `gpt-oss-20b` model page says the model is built for local deployment, uses only 3.6B active parameters, and supports 131k context, so slowdowns are more likely tied to context size, quantization, or offload settings than raw feasibility.
• –For a 16GB VRAM laptop GPU, the model is plausible, but “works” and “feels snappy” are very different bars.
• –This post is a good signal that LM Studio could benefit from one end-to-end setup page for Anthropic/Claude users.