Control your spend
Because it’s your key and your Anthropic account, you hold the levers. There are four of them — from a hard ceiling on the account down to per-agent tuning.
Set spend limits in the Claude Platform
Your first and firmest lever lives where your key does. In the Claude Platform’s billing settings you can set spend limits and usage alerts — a ceiling on what the account can spend, and an early warning as you approach it. Because Anthropic bills you directly, these limits apply to everything your key does, copadre included.
Pick a cost-appropriate model per agent
Every agent in copadre chooses its own model — so match the model to the job: a fast, cheap model for simple FAQ or routing agents, a more capable one for more complex tasks. There’s no reason a simple router should run on your most expensive model.
Enable prompt caching
Enable prompt caching for an agent and copadre reuses its instructions between turns rather than resending them every time — so on longer, multi-turn conversations you pay less for the same chat.
It’s a per-agent toggle, off by default: while you’re still building and testing an agent there’s nothing settled to cache, so you switch it on once the agent is ready. It’s not recommended for concierges — they usually handle a single turn before handing the visitor to a specialist, so there’s no follow-up turn for the cache to save on.
Lean on copadre’s built-in protections
Finally, copadre itself guards your bill: domain locking, abuse blocking and a configurable turn limit prevent inference tokens from being wasted on low quality interactions — other sites can’t use your key, abusive visitors get cut off, and no single conversation can run away with your budget.
Using your own API key means you pay for inference direct. copadre never marks up — or even sees — your AI bill.