DEV Community

dtzp555-max
dtzp555-max

Posted on • Originally published at github.com

Run Claude Opus 4.6 as an OpenAI-compatible API using your Pro/Max subscription ($0 extra)

#claude #openai #proxy #ai

TL;DR

If you're already paying $20-100/month for Claude Pro or Max, openclaw-claude-proxy (ocp) lets you use that subscription through any OpenAI-compatible app — no per-token billing, no API keys.

  • 👉 GitHub: dtzp555-max/openclaw-claude-proxy
  • ⭐ Supports: Claude Opus 4.6, Sonnet 4.6, Haiku 4.5
  • 🔌 Compatible with: LangChain, Open WebUI, LlamaIndex, Cursor, Continue.dev, and anything speaking OpenAI

The Problem

You have Claude Max ($100/mo). You want to use Opus 4.6 in your local agent setup.

Your options used to be:

  • Use the official API → pay $15/M tokens on top of your subscription
  • Manually copy-paste into claude.ai → not practical

What ocp Does 

Your app (OpenAI SDK / LangChain / curl)
    ↓  POST /v1/chat/completions
ocp running at localhost:3456
    ↓  CLI worker pool
claude CLI (your OAuth session / subscription)
    ↓
Anthropic API (billed to your subscription, not per-token)
Enter fullscreen mode Exit fullscreen mode

Cost Comparison

Usage Direct API ocp (Pro $20/mo) ocp (Max $100/mo)
100 Opus requests (2K tokens each) ~$3/run $0 extra $0 extra
Monthly heavy usage (500 requests) $15+ included included
Multiple agents sharing multiply one subscription one subscription

Setup (2 commands) 

git clone https://github.com/dtzp555-max/openclaw-claude-proxy
cd openclaw-claude-proxy
node setup.mjs
Enter fullscreen mode Exit fullscreen mode

That's it. The proxy:

  • Starts on port 3456
  • Auto-installs as a launchd service (macOS) or systemd service (Linux)
  • Restarts automatically if it crashes

Works with Any OpenAI Client 

# LangChain
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(
    base_url="http://localhost:3456/v1",
    api_key="not-needed",
    model="claude-opus-4-6"
)
Enter fullscreen mode Exit fullscreen mode 
# curl
curl http://localhost:3456/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"claude-opus-4-6","messages":[{"role":"user","content":"Hello"}]}'
Enter fullscreen mode Exit fullscreen mode

Honest Trade-offs

⚠️ Serial execution: The CLI worker pool processes one request at a time. This means higher latency than the direct API.

✅ Pre-warming: ocp pre-warms workers so the first real request doesn't pay cold-start cost.

✅ Best for: Personal tools, local agents, dev workflows where you're the only user.

❌ Not ideal for: High-concurrency production servers.

v1.7.x Changelog

  • v1.7.1: Auto-recovery after network drops, sanitizes env to prevent auth conflicts
  • v1.7.0: Optional Bearer token auth via PROXY_API_KEY

Bonus: Share One Subscription Across Agents

If you run multiple OpenClaw agents (different bots, different personas), they can all hit the same ocp instance and share your single Claude subscription.

GitHub: https://github.com/dtzp555-max/openclaw-claude-proxy

Feedback and PRs welcome!

Top comments (0)