the use case i'm optimizing for is "vibe coding" or "agentic engineering" dev loops where an llm is the one primarily writing and running tests. (in vibium's case, the specific kind of tests are end-to-end checks of the user interface in a web browser.)

we started with claude support, but the exposed interface is "just" an mcp server. it should just work with other tools. i haven't tested with them, though. however, others have, and are posting to linkedin. search for "vibium". we'll see if we can get them posting over here, too!

a coding agent could also ignore the mcp server and use vibium like any other browser automation tool, using the javascript/typescript or python client lib and api. in that case, you'd ask the ai to "write a vibium script to test x, y, and z." it could then write, save, and run that script like any other code.

your choice whether to use mcp or the language-specific libs.

ideally, though, the dream use case is you ask your ai friend, "hey, does my site still work? can people still buy stuff from my web store?" and it answers with authority by using vibium to check. (and you never have to worry about the low-level details of how it did it.)