Benchmarking models against your actual workload
Stop guessing which model performs best on your task. Swap between providers in your Stagehand config and compare results on real inputs, in your real environment.
One API key. Every model. Zero friction.
Stop managing credentials for a dozen providers, access all of them with a single Browserbase API key.
A unified model gateway, built into Stagehand.
Model Gateway sits between your Stagehand agents and every model they need.
Anthropic, OpenAI, Gemini, one config, one key, one place to manage it all.
No separate accounts. No rate limit negotiations. No switching overhead when you want to try a different model. It's model access the way it should have always worked.
Every model switch used to be a project. Now it's a config change.
Today, testing a new model means creating an account, generating an API key, wiring it into your environment, figuring out rate limits, and repeating that process every time the landscape shifts. That overhead compounds fast when you're running agents in production.
Model Gateway eliminates all of it. Swap models in a single line. Experiment without friction. Scale without negotiating rate limits from scratch every time a better model ships.
One key, every model
Access Anthropic, OpenAI, Gemini, and more models through your existing Browserbase API key.
Experiment without overhead
Switch models in your Stagehand config without touching credentials or account
Scale without the procurement headache.
Rate limits are handled at the platform level, not yours.
If you're building with Stagehand, this is already for you.
Whether you’re prototyping a new agent or running workflows at scale, Model Gateway fits where you already work. No new SDK. No new authentication flow. If you have a Browserbase API key, you’re already in.
Running production agents without key sprawl
One team, one key, all models. Stop asking teammates to share credentials or maintain separate `.env` files for every provider you've ever tried.
Staying current without migrations
When a better model ships, you update one line. Your agent infrastructure stays exactly the same.
