Latest news, technical deep dives, and automation guides.
Owl Browser v1.2.0 introduces a render mode built for AI agents instead of human eyes. The page comes back as a compact, handle-addressable view, not a screenshot. On a Hacker News front page that is 753 structured tokens an agent can click, instead of a roughly 1,365-token image it has to decode and guess coordinates on.
An autonomous AI agent named Steffi played 35 moves against a Master-level bot on chess.com and won by checkmate, without taking a single screenshot, calling a vision model, or running any OCR. Here's how Owl Browser's browser_get_page_map made it possible.
Owl Browser 1.1.0 takes our stealth and performance to the next level. We now build CEF directly from Chromium source, with Owl-specific patches baked into the engine.
Owl Browser now supports WebMCP, the W3C draft standard that lets web pages declare callable tools for AI agents. Navigate to a page, discover its tools, call them directly. No DOM scraping required.
Your AI agent can now be a verified bot. Owl Browser 1.0.10 ships native Web Bot Auth (RFC 9421) with Ed25519 request signing, Cloudflare Signed Agent registration, and automatic User-Agent switching.
Our new browser_get_page_map tool radically improves how text-based LLMs browse the internet by extracting an interactive map of the DOM, skipping the noise.