It allows you to build agents that can navigate sites, fill out forms, and extract information. It is currently at 90.4% on the Web Voyager benchmark.

With this library, you can: - Run a browser agent suite in your local machine with one command- npx browserable. - Integrate into existing code with REST APIs/ SDKs. - Self-host the browser agent and scale without being tied to any single browser/ browser infrastructure/ LLM provider.

It comes with a UI to run a task and an admin interface to trace, track and debug LLM calls (and costs). Works with multiple browser providers (Steel, Hyperbrowser, Browserbase) and concurrency restrictions.

We built Browserable because we enjoyed building AI agents to automate tasks for our B2B SaaS startup, but we were frustrated at how hard it was to set up and maintain reliable automations even for simple tasks since the groundwork did not exist. We had to integrate with multiple browser providers with various concurrency restrictions, multiple LLM providers with various rate limits and there was no easy way to tie all the logs back to one single task to analyze the accuracy, latency, and cost of the implementation.

We are hoping to make it easier to build, test and improve agents without having to write all of the infrastructure from scratch.

We would love to know what you are building and if this library would be useful, or how we could make it more useful for you!