Launching today
Capstop
A hard dollar cap for every AI agent run
2 followers
A hard dollar cap for every AI agent run
2 followers
A local proxy between your agent and OpenAI. Give each run a dollar cap; when the next call could cross it, your code gets a 402 and the call never goes out. Key goes only to OpenAI. Open source, early, OpenAI chat calls only, tested on Linux.
Free
Launch Team
AWS Startup AdvisorPersonalized AWS guidance that meets you where you’re already building
Promoted
Hi, I'm Abhiram. I build agents, and provider spend limits are monthly while a runaway loop isn't. Capstop caps a single run: calls go through until the next one could cross the cap, then your code gets a 402 and can stop cleanly.
It's early. It handles OpenAI chat completions only, it's tested on Linux only, and it hasn't been checked against a real invoice yet. You can try it in a couple of minutes with no key, against a built-in fake server: clone the repo and run `./run.sh setup` then `./run.sh fake`. The README has the rest.
I'd really like to hear which provider you'd want next, and whether per-run caps match how you work.