Agents are good at the calls and bad at the gaps between them; giving the model an explicit way to wait is what turns a retry loop or a rate-limited sequence into something that finishes.
One tool that does nothing but wait. It exists for the gap between operations: pacing API calls, or giving an eventually consistent system time to settle before the next check.
- `sleep` — waits for a duration in milliseconds, then returns
Clone the repository and `npm install`, then `npm run build`; the client runs the built `index.js` with node. One thing will bite you: the client's own `timeout` setting must be larger than the longest sleep you plan to ask for, or the call fails instead of waiting — the sample config sets `timeout` to 300, so raise it to sit above your longest pause.
