Persistent agent teams without giving them the keys. Puffling gives agent crews memory, skill drafts, sanity checks, and bounded work packets while keeping shell, wallet, publishing, and payout actions review-gated.
A safer way to make agent teams persistent: they remember, draft, inspect, and queue work, but high-blast-radius actions stay behind operator approval.
See the shape of Puffling without buying anything: what it can remember, what it can queue, and what stays review-gated.
Start here if you are curious or comparing agent operating systems. It gives you a clean tour and a safe prompt to hand to your own agent.
Give agents persistent memory, reusable skill drafts, and a bounded task queue without handing them keys.
Choose this when you already have agents doing useful work and want them to remember, draft, and queue without getting command authority.
Operate a named agent crew with review gates, dashboards, weekly findings, and safe handoffs.
Choose this when you want a managed crew, recurring reports, and operator handoffs instead of a pile of endpoints to wire yourself.
Puffling packages the ClawdChat learning loop into a product customers can understand.
truefalsedisabled_without_approved_sandboxfalsequeue_onlyoperator_approved_only