Original Summary

For the past few months I've been losing hours every week to the same browser tasks: applying to jobs, chasing client updates, drafting replies to Jira tickets. All of it on sites with no usable API, so Zapier-type tools couldn't touch it. With tools like n8n, zapier. I never use 98% of their nodes, and the configuration part is hectic for a lazy a** person like me. So i built one for myself with only few nodes that actually mattered to me and no configurations, just signup and automate. I tried the fully autonomous browser agents. They worked maybe half the time and I could never see what they'd actually done, so I didn't trust them with anything that mattered. So I built the middle ground and ended up using it daily. It's called Brovio. How it works: you drag nodes onto a canvas and connect them. Seven nodes total: Start - every run begins here Open URL - navigates the cloud browser to a page Act - one atomic action: click, type, select Observe - inspects what's on the page before you act on it Extract - pulls structured data off the page against a schema Agent - hand it a goal and let the AI drive the browser itself Send Email -delivers the result when the run finishes Every node publishes named outputs, so you write {{ nodeId.title }} in a later node's field and the value flows through. Extract → Send Email is two nodes and zero glue code. The part I care about most: while it runs you get live logs and a structured output panel under the canvas, and a full session replay when it finishes. You can see exactly what the browser did. The canvas is also multiplayer, so two people can build the same workflow at once. Stack: Next.js 16, React 19, React Flow for the canvas, Liveblocks for multiplayer, Trigger.dev for background runs, Stagehand on Browserbase for the browser work, Postgres + Drizzle. Honest limitations , since I'd want to know these: Self-hosting isn't dependency-free. Clerk (auth), Liveblocks (multiplayer) and Browserbase (cloud browsers) can't be self-hosted — all have free tiers. Trigger.dev can be self-hosted or used via cloud. Postgres runs in the Compose stack. No branching or conditionals yet. Workflows are linear. No scheduling or webhook triggers yet — you press Run. The Agent node is the least predictable part, by nature. I use deterministic nodes for anything important and the Agent node only where the task is genuinely fuzzy. AGPL-3.0. Self-host it with Docker and one setup script, free, no feature gates. There's a hosted version too if you don't want to run infra. It's the first open-source thing I've built that actually works and that I use every day, which is why I open-sourced it rather than keeping it to myself. Repo: https://github.com/saksham-dev0/Brovio.cc Live: https://brovio-cc.vercel.app Happy to answer anything about the architecture — especially how runs execute, that took the longest to get right. And if you've tried building browser automation befor

  • 情报分类：服务器与云资源
  • 分类依据：内容涉及服务器、云资源或网络线路
  • 信息来源：Reddit · SideProject
  • 发布时间：2026/9/23 11:04:30