vs
Keep the OpenAI API. Change what it costs.
The OpenAI API meters every token, first-party. ProxyLLM sits in front of it: OpenAI-bound bulk runs through Codex Hosted on your own ChatGPT subscription at a flat monthly rate, and your API key stays configured as the fallback. Same OpenAI-compatible calls, a different bill.
Free for 7 days, then $129/month. Bring your own model keys. No inference markup.
ProxyLLM vs OpenAI API, line by line.
At a glance
ProxyLLM
What the OpenAI API does best.
Be clear-eyed about it, because you are not dropping it. The OpenAI API is the source of truth for OpenAI models: the newest models the day they ship, the highest rate limits, the Batch API discount, and first-party uptime and enterprise agreements. If a workload needs a model that came out this morning or a contractual SLA, that comes from OpenAI directly. Nothing here argues otherwise. The argument is about what most of your volume should cost while you keep it.
What ProxyLLM adds in front of it.
ProxyLLM is an OpenAI-compatible endpoint. You set one base URL, use a ProxyLLM key, and keep your app code exactly as it is. From there:
A flat lane for the bulk
OpenAI-bound work runs through Codex Hosted on your ChatGPT subscription, at a flat monthly cost instead of per token.
Your API key as fallback
When a request needs the direct API, ProxyLLM falls through to your OpenAI key on the same routing key. Nothing hard-breaks.
Scoped keys and hard budgets
Each client, app, or workflow gets its own key and monthly cap. One runaway loop hits its ceiling, not your whole bill.
One endpoint, no markup
Every provider behind one OpenAI-compatible surface, one request log, one cost view. Flat SaaS, no per-token inference margin.
Most teams run both.
Use the OpenAI API directly when you need the newest model immediately, the highest rate limits, or a first-party contract. Put ProxyLLM in front when you are an agency or platform running steady OpenAI volume across clients and you want that volume on a flat subscription, with per-client budget caps and one auditable gateway. The flat lane carries the bulk; the direct key carries the edges. A meter organizes your spend. ProxyLLM changes the price of the OpenAI share of it.
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.
Put your OpenAI volume on a flat plan.
Keep the API for the edges. Run the bulk through Codex Hosted on your ChatGPT subscription, with a hard budget around every client and one endpoint in front of every provider.