capit caps at the provider-key layer — where do you see the boundary for spend that doesn't go through those keys?

Hi — capit's positioning ("manage spending independent of the broken tool and broken agent", per-agent/per-provider cost capping via key management) sits close to a problem I'm working on, so I'd rather compare boundaries than pitch anything.

capit's cap lives at the provider-key layer, which cleanly covers LLM inference bills. The cases I keep wondering about: an agent buying APIs, triggering cloud resources, or paying via x402 — spend that doesn't flow through your keys. And above the cap itself: do you see a need for human approval above a threshold, versioned/auditable policy changes, or defense against an agent splitting one big run into many small ones to stay under a cap?

I'm building a rail-agnostic authorization layer (one policy across payment rails, ALLOW/APPROVAL/DENY, tamper-evident audit) — not to sell anything, but because talking to the team that already solved one slice is the fastest way to learn where the real boundary is. If key-level capping is enough for your users, that's a useful answer too.

No files, tests, or entry points are identified in the issue. First clarify whether the project should expand beyond provider-key cost capping and define a specific, testable change; completion would require an agreed scope and acceptance criteria.