Owl Browser

Owl Browser is a high-performance, self-hosted Chromium and CEF engine designed for automated web scraping and agentic workflows. Instead of heavy screenshots or noisy DOM trees, its OwlMark renderer presents pages as compact, handle-addressable views (like b3 or x2), enabling any model, including text-only models, to operate real web interfaces reliably. In published reproducible benchmarks across six browser tasks, Owl Browser MCP finished each task with a median of 38,590 tokens, compared to 71,437 for Chrome DevTools MCP and 73,420 for Playwright MCP, all maintaining 100% correctness. Key capabilities: - Built-in MCP server running straight from POST /mcp with selectable tool profiles. - Native natural language automation (browser_nla) executing task loops inside the browser engine. - Source-level Chromium fingerprint spoofing across 27 C++ modules and 31 engine patches. - Up to 256 isolated parallel contexts with sub-12ms cold start. - On-device vision engine running Qwen3-VL-2B for local CAPTCHA solving. - Drop-in Playwright and Puppeteer migration pathways. Owl Browser is commercial software with a 14-day $0.99 trial. The free companion build, owl-light, provides a Playwright-compatible drop-in without commercial license requirements (not open source).