Claude, from the OpenAI client you already have.
No Anthropic SDK and no second integration. Request anthropic/claude-3.5-sonnet from the same client as your OpenAI traffic. One request log, one cost dashboard, both providers.
Free for 7 days, then $129/month. Bring your own model keys. No inference markup.
Three steps to connect.
Add an OpenRouter key
Claude runs through OpenRouter, so one key covers Anthropic, Google, and Meta models. Paste it in the ProxyLLM dashboard; it is encrypted with AES-256-GCM.
Point your SDK
Set the base URL to https://api.proxyllm.ai/v1 with your ProxyLLM key as the bearer. No Anthropic SDK, no second client library.
Request anthropic/ models
anthropic/claude-3.5-sonnet, anthropic/claude-3.5-haiku, anthropic/claude-3-opus. The slash in the model name tells the gateway to dispatch via OpenRouter.
Same client, different model name.
The anthropic/ prefix sends the request through your OpenRouter key.
from openai import OpenAI
client = OpenAI(
base_url="https://api.proxyllm.ai/v1",
api_key="pk_live_…", # your ProxyLLM key
)
r = client.chat.completions.create(
model="anthropic/claude-3.5-sonnet",
messages=[{"role": "user", "content": "Review this diff."}],
)Run your AI workloads on your ChatGPT subscription.
ProxyLLM runs OpenAI's Codex for you, signed in with your own ChatGPT account. Your apps call one OpenAI-compatible endpoint and the work bills to your flat plan instead of per-token API pricing.
Run Claude Code inside your apps.
Claude the model runs through your OpenRouter key, above. Claude Code, the agent, is a separate lane: your apps can call it from the same endpoint by requesting bridge/claude-code. It runs on a small bridge on a server you own, next to your own Claude sign-in.
# On a Linux box you own, once:
npx proxyllm-bridge setup
# Then, from any app, through your ProxyLLM key:
curl https://api.proxyllm.ai/v1/chat/completions \
-H "Authorization: Bearer pk_live_…" \
-H "Content-Type: application/json" \
-d '{
"model": "bridge/claude-code",
"messages": [{"role": "user", "content": "Refactor this module."}]
}'Your server
The bridge is a small service on a Linux box you own, or one the dashboard provisions in your own cloud account. Your machine, your bill.
Your sign-in
It runs next to your own Claude Code sign-in and wraps its programmatic mode, claude -p, for your own testing. ProxyLLM stores only the bridge URL, token, and health, never your Claude credentials.
One endpoint
Your apps request bridge/claude-code through a routing key, from the same OpenAI-compatible client. A failing bridge falls through to the next lane on the key.
Anthropic does not permit third-party services to route through Claude subscription credentials, so ProxyLLM never hosts Claude Code and your sign-in never touches our servers. The bridge wraps claude -p for testing on infrastructure you own; you stay responsible for your cloud account, your sign-in, and Anthropic's terms. Set up the bridge.
Add Claude without adding a codepath.
Claude traffic runs on your own OpenRouter key with no inference markup. OpenAI-bound work has its own lane: Codex Hosted serves it from your flat ChatGPT subscription.