1 comments

  • markpolymorph 6 hours ago ago

    Hey HN, built this because I wanted an AI assistant that actually does things instead of describing what it would do.

    Under the hood it's OpenClaw (open-source agent framework) with a real Chromium browser, Brave Search API, and Resend for email. The AI orchestrates these in a loop to complete multi-step tasks.

    Every user gets an isolated session. Built-in rate limiting, RLS on all DB tables, tools locked down so users can't access the server filesystem.

    It's uncensored — presents all sides of topics without refusing or hedging. I think that's a meaningful differentiator vs ChatGPT/Claude/Gemini which all have heavy content filtering.

    Rough spots: complex multi-step tasks need occasional retries, heavy JS sites can trip up the browser. Still early.

    Free to try, no setup: cerebro.vip

    Happy to answer technical questions about the architecture.