Vibium Roadmap

September 8, 2026 · View on GitHub

Things we'd like to work on. Priorities will follow what people need.

Right now

Make browser startup, shutdown, and error handling more reliable.
Finish real-world testing with Anthropic, Gemini, and local models.
Make setup smooth in Codex and Claude.
Evaluate how well independent checks catch mistakes, using the same model or a different one.

The main remaining piece of AI-powered locators: ask for an element and get one back. Something like vibe.find("the blue submit button").

We still need to work out ambiguous matches and how prompts fit alongside CSS selectors and semantic locators.

Help agents remember pages and paths between sessions, so they don't keep rediscovering the same things.

If that proves useful, give people a way to browse and edit the memory too. There's an early UI prototype.

Edge and Safari, as people need them. Video recording in more browsers, as their protocol support allows.

Bring official C# support to Vibium. There's already an experimental community implementation to learn from.

Container images and recipes for running Vibium in CI. Further out: hosted Run and Check, cloud browsers for those operations, and PR integration.

Support more Playwright trace versions. Explore saving a new Check's evidence alongside an existing recording, while keeping the original intact.

Open an issue. Tell us what you're trying to do and what's getting in the way.