Reference
Gateway (coming soon)
The gateway is for teams that cannot change code, use a language without an SDK, or want limits that stop calls.
How it will work
You run the Zexl gateway, an open-source LiteLLM proxy with a Zexl plug-in, in your own cloud. Your apps change one base URL to point at it. It forwards calls to your providers with your own keys, and sends Zexl the same metadata the SDK sends. Prompts stay in your network.
text
Your app ──▶ Zexl gateway (in your cloud) ──▶ OpenAI, Anthropic, Bedrock, …
└──▶ call metadata only ──▶ ZexlWhat it adds
| Gateway | SDK | |
|---|---|---|
| Spend, waste patterns, forecast | Yes | Yes |
| Any language, no code change | Yes | Python only |
| More than 100 providers | Yes | OpenAI and Anthropic |
| Stop a call (budget limit, max cost per run, allowed models) | Yes, off by default | No, alerts only |
| Retries, fallbacks, caching | Yes | No |
| Extra network hop | A few milliseconds | None |
Labels come from request headers (x-zexl-feature, x-zexl-run, x-zexl-customer) or from the gateway key's name.
Want it early? Tell us which providers and limits you need.