Vidur
A browser assistant that does more than summarize. Vidur finds what is actionable on a page, suggests the next question, and helps you move from reading to doing.
Turn browsing into progress.
Most page summarizers compress information. Vidur is designed around the decision that comes after reading: what matters, what should happen next, and what question would make the research sharper.
It combines a browser side panel, URL-based conversation history, site-specific skills, model choice, voice, sharing, and a confirmed action flow that lets a vision model locate an element before a click is executed.
A research copilot with memory.
Structured understanding
Summaries, actionable items, red flags, and follow-up questions are generated from the page context.
Reusable skills
Job postings, products, LinkedIn profiles, PDFs, and search results each get a tailored analysis path.
Natural interaction
Voice input, read-aloud output, copy, sharing, and per-page conversation threads keep the flow lightweight.
Confirmed actions
When Vidur proposes a click, the user sees the screenshot and target before anything is executed.
Designed around trust.
Vidur uses Chrome's activeTab and scripting permissions instead of broad page access. Sensitive actions use a screenshot, target marker, and explicit confirmation before a real click. Conversations are mirrored to a local web viewer on port 4300, while API keys remain in extension storage.