Claude Code

Claude Code is Anthropic's agentic coding CLI. It edits files, runs shell commands, and uses tools to complete multi-step development tasks.

Qubrid serverless endpoints are OpenAI-compatible (/v1/chat/completions). Claude Code speaks the Anthropic Messages API (/v1/messages). You cannot point Claude Code directly at Qubrid-you need a small translation proxy in the middle.

This guide walks through that setup end to end so you can run Qubrid OSS models (for example GPT OSS 120B or Qwen3 Coder) from Claude Code.

How it works

Claude Code  →  Translation proxy  →  Qubrid API
               /v1/messages            /v1/chat/completions

1. Claude Code sends Anthropic-format requests to the proxy.
2. The proxy converts them to OpenAI-format chat completions.
3. Qubrid runs your chosen OSS model and returns the response.
4. The proxy converts the response back to Anthropic format for Claude Code.

Claude Code relies heavily on tool calling (function use). Pick a model that supports tools on Qubrid-agentic coding models such as GPT OSS 120B and Qwen3 Coder are good starting points.

Install Claude Code

Claude Code supports a native installer, Homebrew, and npm. See the Claude Code installation docs for all options.

Native install (recommended):

curl -fsSL https://claude.ai/install.sh | bash

Homebrew (macOS):

brew install --cask claude-code

After installation, open a terminal in your project and run:

claude

You do not need an Anthropic subscription for this guide-the proxy routes requests to Qubrid instead of api.anthropic.com.

Verify Qubrid works

Before configuring the proxy, confirm your Qubrid API key and model ID work.

export QUBRID_API_KEY="your-qubrid-api-key"

curl -s https://platform.qubrid.com/v1/chat/completions \
  -H "Authorization: Bearer $QUBRID_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-oss-120b",
    "messages": [{"role": "user", "content": "Say hi in one sentence."}],
    "max_tokens": 50
  }'

Set up the translation proxy

Choose one of the options below. Option A is the fastest way to get started. Option B is better for teams that want centralized routing, logging, or multiple model aliases.

Option A: claude-code-proxy (recommended)

claude-code-proxy is a lightweight proxy that converts Claude Code's Anthropic requests to OpenAI-compatible calls.

git clone https://github.com/fuergaosi233/claude-code-proxy.git
cd claude-code-proxy
pip install -r requirements.txt

cp .env.example .env

# Qubrid credentials
OPENAI_API_KEY="your-qubrid-api-key"
OPENAI_BASE_URL="https://platform.qubrid.com/v1"
 
# Map Claude Code model tiers to Qubrid OSS models
BIG_MODEL="openai/gpt-oss-120b"
MIDDLE_MODEL="openai/gpt-oss-120b"
SMALL_MODEL="Qwen/Qwen3-Coder-Next"
 
# Proxy listens on port 8082 by default
HOST="127.0.0.1"
PORT="8082"

python start_proxy.py

Option B: LiteLLM (teams and gateways)

LiteLLM is a gateway Anthropic documents for enterprise Claude Code deployments. It can route Claude model names to Qubrid OSS models.

pip install litellm

model_list:
  - model_name: claude-sonnet-4-6
    litellm_params:
      model: openai/openai/gpt-oss-120b
      api_base: https://platform.qubrid.com/v1
      api_key: os.environ/QUBRID_API_KEY
 
  - model_name: claude-haiku-4-5
    litellm_params:
      model: openai/Qwen/Qwen3-Coder-Next
      api_base: https://platform.qubrid.com/v1
      api_key: os.environ/QUBRID_API_KEY

export QUBRID_API_KEY="your-qubrid-api-key"
litellm --config config.yaml --port 4000

Configure Claude Code

Point Claude Code at your local proxy instead of Anthropic's API.

{
  "$schema": "https://json.schemastore.org/claude-code-settings.json",
  "env": {
    "ANTHROPIC_BASE_URL": "http://localhost:8082",
    "ANTHROPIC_API_KEY": "any-value",
    "DISABLE_TELEMETRY": "1",
    "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1",
    "ENABLE_TOOL_SEARCH": "false"
  }
}

{
  "$schema": "https://json.schemastore.org/claude-code-settings.json",
  "env": {
    "ANTHROPIC_BASE_URL": "http://localhost:4000",
    "ANTHROPIC_AUTH_TOKEN": "sk-litellm-static-key",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4-6",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-haiku-4-5",
    "DISABLE_TELEMETRY": "1"
  }
}

cd your-project
claude

Explain what this repository does in three bullet points.

Recommended Qubrid OSS models for coding

These models on Qubrid are well suited to agentic coding workflows:

Copy the exact model ID from the AI Playground or the model detail page.

Verify the full stack

Run these checks if something does not work.

1. Qubrid responds directly (see Verify Qubrid works above).

2. Proxy responds to Anthropic-format requests:

curl -s http://localhost:8082/v1/messages \
  -H "x-api-key: test" \
  -H "Content-Type: application/json" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 30,
    "messages": [{"role": "user", "content": "Say hi"}]
  }'

3. Claude Code sees the proxy:

/status

4. Usage appears on Qubrid:

Open Inference Logs after a Claude Code session to confirm requests reached Qubrid.

Things you can do with Claude Code + Qubrid

Once setup is complete, try these workflows:

• Explore a codebase - Ask Claude Code to summarize architecture, find entry points, or trace how a feature works.
• Fix bugs - Paste an error or describe a symptom; Claude Code can search the repo and propose a fix.
• Write tests - claude "write tests for the auth module, run them, and fix failures".
• Refactor safely - Describe the change you want across multiple files and review the diffs before accepting.
• Automate chores - Lint fixes, dependency updates, commit messages, and small PR prep.

For more commands and flags, see the Claude Code documentation.

Monitor usage on Qubrid

Keep an eye on your usage and costs on the Qubrid platform.

• Review Inference Logs to see request history and debug issues.
• Check your credit balance before long agentic sessions-Claude Code can send many requests per task.
• Rotate API keys if you suspect a key was exposed.

Configuration reference

Claude Code's built-in model discovery (CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY) only lists models whose IDs start with claude or anthropic. Qubrid OSS model IDs will not appear in /model automatically-use the proxy's model mapping (BIG_MODEL, LiteLLM model_list, etc.) instead.

Frequently Asked Questions (FAQ)