TryCase

TryCase is a testing infrastructure tool built for developers who work with coding agents. It gives any AI coding assistant — Claude, Codex, Cursor, or others — a disposable Linux desktop to run your application, click through it like a real user, and return concrete proof of what happened. Instead of relying on the agent's word that the code works, you get a video recording, screenshots, logs, and any artifacts generated during the test.

Using TryCase is straightforward. You install its skills into your coding agent with a single npx command: `npx skills add bencsn/trycase-skills`. After that, you just ask the agent to test something. The agent spins up a Linux environment with terminal and browser access, uploads your code, opens the app, and runs through the flow you requested. Everything is recorded. When the test finishes, the environment is destroyed — no leftover files or running processes. The agent returns the evidence alongside its answer, so you can review exactly what happened.

TryCase is most useful for developers who want to ship faster without sacrificing confidence. If you use AI agents to write code or fix bugs, you can ask the agent to "verify end to end that it is fixed in TryCase" before merging a pull request. It also helps with reproducing bugs in a clean environment, reviewing branches from a user perspective, and testing dependency upgrades for visual or runtime regressions. The tool is simple — install once, then prompt for proof on every task.