🚧 In development — reserving early access (not live yet)
Stop the $6,000 overnight LLM bill.
A hard kill-switch for your OpenAI / Anthropic API keys.
Not another dashboard — an inline cap that actually blocks the request when your agent loops or a key leaks.
Team plan $29/mo (unlimited seats) · free tier for 1 key · no enterprise sales
Why this exists
OpenAI's "budget limit" is just an email now. It keeps billing after you cross it — there is no true hard stop.
(HN 45589628 / OpenAI community 1193635)
A runaway agent or leaked key can burn $1,800–$6,000 in two days before you even notice.
(documented agent budget overruns; leaked-key studies)
Retry loops fire 3–10 calls/sec. $50/day quietly becomes $500/hour, and nothing stops it.
(getcapped.app and others surface the same pain)
What makes it different
Enforcement, not observation. Dashboards (Langfuse, Helicone) show you the bill after it happened.
Self-hosted proxies (LiteLLM) can enforce — but you have to run Postgres, Redis, and the proxy yourself, and own the outages.
This sits inline and hard-fails the request at your cap — zero ops, no self-hosting. Point your
base_url at it, set a daily/monthly cap per key, done in ~5 minutes. Start in monitor-only, flip to hard-block when you trust it.
Honest positioning: if you already run LiteLLM and enjoy operating it, you don't need this. This is for teams spending $100+/mo on LLM APIs who want the cap without running the infra.
Reserve early access
Two ways in. Both get you priority access at ship time.
↑ Capture endpoint not wired yet (validation prep). Will be enabled before launch.
Straight with you: this is in development. Right now it's an early-access reservation, not a live product.
We don't promise any unshipped feature. If enough teams pre-book, we build it and you get first access.
We'll share progress in public (build-in-public). Not affiliated with OpenAI or Anthropic.