Appearance
Browserable
Category: Framework
License: MIT (open-source, self-hostable)
Pricing: Free. No paid tiers. Self-host your own instance; you supply the LLM API keys (Gemini, OpenAI, or Claude) and a remote browser provider (Hyperbrowser or Steel).
Official site: https://www.browserable.ai
Docs: https://docs.browserable.ai
GitHub: https://github.com/browserable/browserable
Community: Discord
What it is
Browserable is an open-source browser automation framework for AI agents, written in JavaScript. It exposes a self-hosted task runner (Docker-based) that an agent hits via a JS SDK or REST API to perform real browser tasks: navigate pages, fill forms, extract data, and complete multi-step flows.
Benchmark: 90.4% on the Web Voyager benchmark, which measures real-world web task completion.
Key design decisions:
- Self-hosted only. You run the server; no data leaves your infrastructure.
- Requires Docker and Docker Compose to launch the server stack.
- Requires at least one LLM provider key and one remote browser provider key before tasks can run.
- The JS SDK (
browserable-js) is a thin, type-safe wrapper over the REST API.
Service ports (after startup):
| Port | Service |
|---|---|
| 2001 | UI (settings, task dashboard) |
| 2002 | Docs server |
| 2003 | Tasks API |
| 3300 | MongoDB Express |
| 9001 | MinIO Console |
Safety and ToS notes
- MIT license. No usage restrictions beyond the license terms.
- You bring your own LLM and browser provider keys. Review the ToS for whichever providers you connect (OpenAI, Gemini, Anthropic, Hyperbrowser, Steel).
- Browserable itself has no rate limits or usage policies beyond your own infrastructure capacity.