> ## Documentation Index
> Fetch the complete documentation index at: https://docs.muna.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Claude Code

> Run Claude Code on open models served by Muna.

[Claude Code](https://code.claude.com) talks to the Anthropic Messages API, which Muna implements.
Point it at Muna with a few environment variables.

## Configuring Claude Code

```bash theme={null}
export ANTHROPIC_BASE_URL="https://inference.muna.ai"
export ANTHROPIC_AUTH_TOKEN="$MUNA_API_KEY"
export ANTHROPIC_MODEL="@qwen/qwen-3.8-27b"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="@qwen/qwen-3.8-27b"
```

Then start Claude Code in your project directory:

```bash theme={null}
claude
```

<Note>
  Claude Code sends background work, like summarizing a conversation, to a smaller "Haiku" model.
  Muna does not serve Claude models, so set `ANTHROPIC_DEFAULT_HAIKU_MODEL` to a Muna model too.
  Otherwise those requests fail.
</Note>

To keep the settings across sessions, add them to the `env` block of `~/.claude/settings.json`:

```json theme={null}
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://inference.muna.ai",
    "ANTHROPIC_AUTH_TOKEN": "<your Muna API key>",
    "ANTHROPIC_MODEL": "@qwen/qwen-3.8-27b",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "@qwen/qwen-3.8-27b"
  }
}
```

<Tip>
  Any chat model in [Models](/inference/models) works. Qwen 3.8 27B is the strongest
  coding model on Muna today; Gemma 4 26B is faster and cheaper for lighter tasks.
</Tip>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.