1. Install Claude Code

Shell
npm install -g @anthropic-ai/claude-code
# Linux / macOS alternative
curl -fsSL https://claude.ai/install.sh | bash

2. Configure ~/.claude/settings.json

JSON
{
  "env": {
    "ANTHROPIC_BASE_URL": "http://127.0.0.1:8313",
    "ANTHROPIC_AUTH_TOKEN": "1",
    "ANTHROPIC_MODEL": "YOUR_MESSAGES_MODEL_ID",
    "API_TIMEOUT_MS": "1200000"
  }
}

Replace the placeholder with an approved gha_... token when ghc-api authentication is enabled. Do not append /v1 to the Anthropic base URL.

3. Select a Messages model

Shell
curl http://127.0.0.1:8313/v1/models/full/

Choose a model whose supported_endpoints includes /v1/messages. Copy the live model ID rather than a version from an old guide. The older ANTHROPIC_SMALL_FAST_MODEL variable is deprecated in current Claude Code documentation.

4. Troubleshoot

  • 401: missing or unapproved token.
  • 404: the base URL probably contains an extra /v1.
  • Unsupported model: the account or organization has not enabled it, or it lacks Messages support.
  • Idle stream: keep SSE keepalive enabled and use a sufficiently long client timeout.

Claude Code settings ↗

Next step

Inspect the models and endpoints available to your account before copying a model ID into client configuration.

Inspect models and parameters →