Web Browsing
Chief can browse the live web. Not search-engine snippets — full pages rendered in a real headless browser.
How It Works
Chief uses Playwright under the hood. When you ask Chief to look something up, it:
- Opens a headless Chromium browser.
- Navigates to the URL (or runs a search and clicks through to the top result).
- Waits for the page to render (including JavaScript-driven content).
- Extracts text, links, tables, or screenshots as needed.
- Reasons over the extracted content and replies.
What You Can Ask
Chief, go to crunchbase.com and find me the latest funding round
for Anthropic. What was the date and amount?Pull the pricing table from stripe.com/pricing and put it in a
sheet I can edit.Watch this Hacker News thread for the next hour and ping me
in #curiosity if anyone posts about Convex.Structured Extraction
When you need data in a specific shape, Chief generates a schema, scrapes against it, and returns clean structured output.
Chief, scrape the team page at example.com/team. For each person,
get name, role, LinkedIn URL, and email if shown. Give me a CSV.Chief returns a CSV file as a Slack upload, ready for download.
Anti-Bot Walls
Some sites block headless browsers. Chief will detect a Cloudflare/Turnstile wall and switch to a stealth mode browser fingerprint that handles most public-site protections. If the page truly requires login, Chief will say so and ask whether you have credentials in your vault.
Privacy
Browsing happens on the agent host, not on your laptop. Chief does not record screenshots or HTML to your dashboard unless you explicitly ask it to attach evidence. Each browser session is destroyed after the task ends — no cookies persist between tasks.
Limitations
- Chief won’t bypass paywalls or login walls without explicit credentials.
- Chief won’t perform actions on third-party sites that move money or send messages without an approval.
- Streaming media (YouTube, video) is not transcribed by the browser tool — use the file processor with a downloaded file instead.