AI agents can pay for things on their own. The hardest part isn’t speed—it’s “who authorized it, and how much can it spend?”

This is what I’ve been thinking about most while researching Kite. Kite’s Agent Passport lets users approve a “spending session” with a passkey: set a per-transaction limit, a total session budget, and an expiration time. The agent can only call paid APIs or place orders within those limits; anything beyond that is blocked. Permissions are programmable and expire—they don’t require handing over your entire wallet.

I believe the real bottleneck for the agent economy is trust: merchants need to verify that an agent has genuine authorization, and users need to know the agent won’t run amok. Kite aims to address identity, spending limits, and settlement within a single infrastructure. That’s the right direction.

But I do have a question: publicly available data doesn’t yet show a clear increase in agent payment volume. What’s worth watching next is the actual number of calls and the scale of settlements—not the buzz.

If you let an AI agent spend money on your behalf, what budget would you give it?

$KITE #KiteBBW