Claude Code

Overview

Anthropic's terminal coding agent. It reads ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN, so an OpenAI-compatible gateway needs a translating proxy in between. LIBRA currently exposes only the OpenAI-style /v1/chat/completions; use it through Claude Code with claude-code-router or another Anthropic-to-OpenAI bridge.

You need a LIBRA API key (libra_sk_...) from the console (API keys > Create key), and the base URL https://<your-host>/v1.

Installation

1. npm install -g @anthropic-ai/claude-code
2. npm install -g @musistudio/claude-code-router (bridge)

Configure LIBRA

~/.claude-code-router/config.json:

{
  "Providers": [{
    "name": "libra",
    "api_base_url": "https://<your-host>/v1/chat/completions",
    "api_key": "libra_sk_...",
    "models": ["libra/smart", "asix-deepseek-v4/deepseek-v4-flash"]
  }],
  "Router": { "default": "libra,libra/smart" }
}

Quick start (up and running in 5 minutes)

1. Save the config above.
2. ccr code starts Claude Code through the router.
3. Ask "Siapa kamu?" - the reply comes from LIBRA. Check Usage in the console.

FAQ

• 404 / timeout? The base URL is missing /v1 or has an extra /chat/completions. Use exactly https://<your-host>/v1.
• Model not found? Model ids are case-sensitive and must match the console Models page, e.g. libra/smart or asix-deepseek-v4/deepseek-v4-flash.
• 401? The key is wrong, disabled or expired. 402? Your organization balance is Rp 0; see Credits.
• 403 model_not_allowed? The key's guardrails block that model. Ask your workspace Admin or pick another model.