It reproduces the bug before it touches the code.
Guess-and-check fixes from a chat window waste your afternoon. Fulminare installs the project, runs it, reproduces the failure, and only then edits the code, then runs the tests again.
Clone my GitHub repo, install dependencies, and run the test suite. For each failing test, reproduce it, find the root cause, fix it, and re-run the tests. Summarize every change and why.
How it does it
- 01Gets your code from GitHub or an upload, and installs dependencies.
- 02Runs the tests or the failing command to see the real error.
- 03Narrows down the cause with targeted reads and greps.
- 04Applies the fix and re-runs until it passes.
Full shell, not suggestions
It has shell, file read/write, patching, glob, and grep on its own sandbox, the same tools you would use.
Your repo stays yours
Every session runs on an isolated sandbox. Connect GitHub and it can push the fix to a branch for you to review.
Questions
What languages does it support?
Anything that runs on Linux: JavaScript, TypeScript, Python, Go, Rust, and more.
Will it push to main?
Only if you tell it to. Ask for a branch and it uses one.
Is Fulminare free?
Yes. The free plan includes 3 live sandboxes, 200 runs / day, 100k tokens / mo, 5 overnight runs / mo — no card required.
How is this different from ChatGPT?
A chatbot tells you how to do the task. Fulminare does it: every session runs on its own cloud machine with a real shell, browser, and file system, and you get the finished files back.
Related
Stop reading answers.
Start getting work done.
Busy people save hours every week when the agent stops advising and starts doing. Point Fulminare at the job. Come back to it finished.