A weekly newsletter on Vibium — the browser automation framework for the AI era.
Recipes across the vibium CLI, the MCP server, and JS / Python / Java clients.
Edge cases from agent-driven runs, and the parts nobody writes down.
Each issue covers one surface of Vibium in depth — a command deep-dive, a real-world recipe, an edge case worth knowing. No fluff, no filler.
Every command in the Vibium CLI, documented with animated terminal demos. Navigate, find, interact, capture, manage state, and wire up AI agents — all from a single binary, all covered in the first issue.
Read Issue No. 01 →
Every MCP tool documented: navigate, find, interact, capture, manage state,
control time. How to wire Vibium into Claude and Cursor, the browser_map
+ @ref pattern that makes agent sessions reliable, and the virtual
clock surface for testing time-dependent behaviour.
Every JS method documented: navigate, find elements, interact, capture dialogs, wait, run page utilities and network capture. The find-once-act-many pattern with element handles, the virtual clock for time-dependent tests, and isolated browser contexts for parallel sessions.
Read Issue No. 03 →The same object model as the JS API — browser, context, page, element — in the language QA engineers actually reach for. Synchronous by default, snake_case throughout, with context managers for dialog and network capture, and the virtual clock for time-dependent tests.
Coming soon
Statically typed, camelCase, with a fluent options builder wherever arguments
are needed. FindOptions, StartOptions,
BrowserContext with try-with-resources teardown, and
Supplier<Boolean> for custom wait conditions.
The full API tour closes here.
They drive the same browser, but they diverge where it matters — HTTP navigation, dialog handling, empty-page text extraction, multi-tab orchestration. 27 paired comparisons, 50+ MCP-only tools, and 14 behaviors that are identical on both. A field guide compiled across 99 QA practice sites.
Coming soonBoth implement MCP. Both automate browsers. But Vibium bets on CSS selectors and @refs while Playwright bets on the accessibility tree. Tool-by-tool comparison: 85 vs 68 tools, reading models, interaction patterns, opt-in caps, and what each server does that the other can't — validated on a live e-commerce AUT.
Coming soonBoth frameworks wait for elements to be ready before acting. But they poll at different intervals, implement stability differently, and expose the timeout knob on very different surfaces. CLI, MCP, JS, Python, Java — each interface has its own story.
Coming soon