Your coding agent says done. Runstory shows what broke.
Runstory builds the app in your commit and runs it in a hosted sandbox. AI agents use it as a person would. Each problem comes back with a script that reproduces it. It works from the command line and through the GitHub App.
Your coding agent says
Done. The list now shows the next task.
Runstory shows
A sample check1 problem found
The video of the sample check, from a hosted sandbox. It is paused at 00:49, the last frame before the board goes blank. F1 is marked at 00:52. The Anthropic agent points at the last open task. The OpenAI agent points at the summary line.
Hosted sandbox
- Install
- Build
- Start
Browserlocalhost:5173
OpenAIAnthropic
F1
The board goes blank when the last task is done
00:52High severity
- Expected
- The board says “Nothing left to do”.
- Verified
- 3 of 3 re-runs failed.
- Reproduce
sh hunt.c1.sh
Your coding agent reads it from .runstory/findings.json
Fixing F1.
Sign in to runstory
Or start in your terminal
curl -fsSL https://runstory.vercel.app/It installs on macOS or Linux. Then run runstory login in your project.
How Runstory works
Each finding is run again
Its script runs 3 more times. Then a blind verifier runs it again.
You keep working
Commits and pushes are never blocked. Uncommitted edits are not part of a check.
Read what source and evidence travel