Show HN: Never Do On-Call Again
After a ton of work, it’s now ready to use. It’s called… well… OnCall, and it’s your AI on-call assistant.
OnCall solves errors thrown from your app, using all the tools and resources you do. It looks at the stack trace and reads source code to understand the error. It sifts through logs, browses the internet, and executes shell commands. Then it codes, commits, and creates a pull request with a nice description.
The performance is surprisingly good. On my limited early testing, it solves most errors perfectly in one try, and makes progress on many it misses. Fixes are isolated on a branch, so if the fix isn’t great, just don’t merge!
It connects via SDK and is triggered automatically when your app throws an exception. When it works well, it’s a fix PR for the error that just lands in your lap around 90 seconds after the error occurs!
Here are different ways to check out OnCall, depending on how busy you are:
[15 seconds] - No login public preview dashboard: https://web.oncallapp.ai/public/demo
[2 minutes] - Video demo: https://youtu.be/OZIoCF2-Hlk
[7 minutes] - Sign up and try it: https://oncallapp.ai
Necessary disclaimers: You should definitely review the code it produces. OnCall connects to your Github and Slack at a minimum, and LLM providers (Google/Anthropic/OpenAI) process part of your code. If you can’t use ChatGPT or Cursor at your work, you probably shouldn’t use OnCall.
Looking forward to hear feedback and ideas!