I created an open source autoscaling AI browser agent
Serverless browser agents with mutual TLS beats running Playwright on one machine.
Distributed, AI-driven browser automation cluster.
Visual diff trees beat screenshot dumping for token efficiency at scale.
Backend engineers, DevOps teams, automation developers
browser-use · Stagehand · Browserbase
Finished working on a browser automation system via natural language and API calls.
This is how it's different than the likes Vercel browser, browser use, stage hand. e.t.c.
Handles identity / proxy management for you.
Handles server management with the aid of github.com/dashn9/serverless-flux enabling serverless browser computing. Spin as much AI browser process as possible. Extremely useful for startups and companies that wants to keep this internal.
Has less memory footprint and much faster (for obvious reasons - which is RUST!!!).
Works extremely well with faster, cheaper models. I'm always actively testing and developing with smaller models and will continue to do so, they are quite underestimated but tremendously capable.
High token efficiency. We don't dump full HTMLs in like browser use, also, I shifted away from screenshot, instead using a visual diff tree which works incredibly well
You just plugin your service account / AMI key (currently supports GCP, AWS) and in less than 30 seconds you are done with the setup. It downscales and upscales based on your demand. Once you are done. auto descales to the min servers you've set.
it's here: https://github.com/dashn9/rusty-browser
Other projects that enable for Rusty function are:
(Serverless management and deployment)[Flux: https://github.com/dashn9/serverless-flux ]
(Browser automation, BiDi and CDP)[Rustenium: https://github.com/dashn9/rustenium ]
(Browser automation, BiDi and CDP)[Rustenium Identity: https://github.com/dashn9/rustenium-identity ]
All contributions and advice and support are appreciated.
Serverless browser agents with mutual TLS beats running Playwright on one machine.
Hands-on distributed ML simulator—gamified learning for tensor parallelism without spinning up clusters.
Browser-based GPU cluster for LLM inference with HTTP API and SSE broker coordination.
Erlang OS nodes clustering directly inside your browser via v86 emulation.
Gives Claude its own machines—VM primitive for AI agents beats hogging your ports and RAM.
REAPI-compatible distributed compute without Docker, etcd, or third-party Raft.