CCAS — Claude Code API Server

An OpenAI-compatible HTTP proxy that routes requests through the Claude Agent SDK, using your Claude Max subscription. No API key required.

Endpoints

Method	Path	Description
POST	/v1/chat/completions	OpenAI-compatible chat completions (streaming + non-streaming)
GET	/v1/models	List available Claude models
POST	/api/chat	Ollama-compatible chat endpoint
GET	/api/tags	Ollama-compatible model list

Setup

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Run

python server.py

Server starts on http://localhost:11435.

Usage

Point any OpenAI-compatible client at http://localhost:11435/v1:

from openai import OpenAI

client = OpenAI(base_url="http://localhost:11435/v1", api_key="unused")
response = client.chat.completions.create(
    model="claude-sonnet-4-6",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Model Aliases

Common OpenAI model names are mapped to Claude equivalents:

OpenAI	Claude
gpt-4, gpt-4o, gpt-4-turbo	claude-sonnet-4-6
gpt-3.5-turbo	claude-haiku-4-5

Dependencies

• claude-agent-sdk — Claude Code Agent SDK
• fastapi — HTTP framework
• uvicorn — ASGI server