[Mr. Browser]
>

Project

Changelog

All notable changes are documented here. Mr. Browser follows Semantic Versioning.

Tip
To upgrade, pull the latest tag and rebuild: git pull && make build && make install. Check the breaking changes section before upgrading across minor versions.

v0.2.0

2026-07-01minor
  • featNew debug command — step through workflows interactively with ENTER/skip/quit prompts.
  • feathover: step type — trigger mouseover/mouseenter events on any element.
  • featupload: step type — set file inputs for upload automation.
  • featextract: step type — save element text to session memory for use in later steps.
  • featreload: step type — reload the current page.
  • featwait.url — wait until the page URL contains a given substring.
  • featassert.element_exists — assert an element matching an intent exists on page.
  • featinspect --target flag — test a specific intent string and see ranked candidates.
  • featPython SDK: hover(), upload(), extract_text(), execute_js(), get_cookies(), set_cookies(), wait_for_url(), assert_url_contains() methods added.
  • featTypeScript SDK: matching methods added.
  • perfDOM Accessibility snapshot caching — 40% faster on pages with repeated intent lookups.
  • fixSelf-heal threshold now correctly defaults to 0.85 (was 0.7 in v0.1.0).
  • fixPassword values are now masked as *** in all logs regardless of field name.

v0.1.0

2026-06-01minor
  • featInitial public release.
  • featCLI: run, screenshot, inspect commands.
  • featYAML steps: open, click, type, scroll, screenshot, wait, assert (text_visible, url_contains).
  • featPython SDK with MrBrowser client and Page context manager.
  • featTypeScript SDK with MrBrowser class and Page.
  • featMemory Engine: SQLite fingerprint store with structural self-healing.
  • featGlobal flags: --headless, --no-sandbox, --db, --log-level, --chromium, --config.
  • featDocker Compose setup with Chromium, volume mounts, and health checks.
  • featDOM Accessibility Tree extraction via Chrome DevTools Protocol.
  • featLocal NLP intent resolution — no cloud, no LLM latency.