We’ve been heads-down for a while, so this feels good to write: Gatelit is now in early access.

The problem

If you’ve shipped anything with LLMs, you know the pattern. You start with one provider — probably OpenAI — and it works great. Then someone wants to try Anthropic for a specific task. Then Google’s latest model looks interesting. Then your team grows and suddenly prompts are scattered across repos, nobody knows what the thing is actually costing, and debugging a hallucination means digging through CloudWatch at 11pm.

Three things always break at scale:

Provider lock-in. Every provider has its own SDK, its own quirks, its own way of handling structured output. Switching means rewriting integration code. Teams punt on it, then regret it.
Zero visibility. You’re blind to what your users are asking, what each request costs, and when the model goes off the rails. It’s the fastest feature you’ll ship and the first one that becomes a black box.
Prompt sprawl. Prompts live in code comments, Notion pages, and that one Slack thread from three months ago. No versioning, no testing, nobody owns them.

We hit all three building our own AI features. So we built something to fix them.

What Gatelit does

Drop our SDK into your backend, point your calls at Gatelit, and it handles the rest:

One API for every provider. Send a request with a model name — Gatelit routes it to the right provider. No provider-specific integration code.
Every request logged. Tokens, latency, cost, the full request and response. You can finally answer “what is this actually costing us?”
Prompts as first-class citizens. Version them, test them in a playground, deploy without a deploy. No more prompt-in-code-comment chaos.
Access control that makes sense. API keys scoped to users, teams, or projects — with granular permissions.

Who it’s for

Engineering teams shipping LLM features in production. If you’re past the prototype stage and starting to feel the friction of managing multiple models, providers, and prompts — that’s where we help.

What’s next

We’re in early access right now. If any of this sounds familiar, join the waitlist and we’ll get you set up. Or browse the docs to see what’s under the hood.

More soon. We’ve got some stuff shipping in the next few weeks that we’re genuinely excited about.