I'd be fine giving an agent $10 to collect some market data. But imagine coming back two hours later and finding out it spent $8 on information I already had.
I'd be annoyed, even if it hadn't broken any spending rules.
I was reading about Kite Passport's spending sessions and trying to figure out how much control I'd actually have over that money.
I can approve a session with my passkey, set a $2 limit for each payment, keep the total budget at $10 and make the permission expire after two hours. I can also choose which assets the agent is allowed to use.
That's pretty useful. I really wouldn't want to approve every little payment while it's trying to finish a task.
One thing in the documentation was worth paying attention to though.
The task summary and payment policy aren't the same thing.
I could write "collect market data" in the task summary, but those words alone don't control what the agent buys. The payment policy sets the actual spending restrictions.
And this is where I still have a problem.
A $2 payment limit can stop a $5 transaction, but it won't stop the agent from making four useless $2 purchases.
It could follow every rule I approved and still waste most of my money.
I can see myself using Kite Passport for smaller tasks because I don't want to sit there approving payments all day. But I'd want to check what the agent actually paid for before trusting it with a bigger budget.
Keeping my money within a limit is good. Spending it on something worth buying is a different matter.
If your agent wasted $8 from a $10 budget without breaking any rules, would you give it another chance or start approving its payments yourself?
@KITE AI $KITE #KiteBBW
I'd be annoyed, even if it hadn't broken any spending rules.
I was reading about Kite Passport's spending sessions and trying to figure out how much control I'd actually have over that money.
I can approve a session with my passkey, set a $2 limit for each payment, keep the total budget at $10 and make the permission expire after two hours. I can also choose which assets the agent is allowed to use.
That's pretty useful. I really wouldn't want to approve every little payment while it's trying to finish a task.
One thing in the documentation was worth paying attention to though.
The task summary and payment policy aren't the same thing.
I could write "collect market data" in the task summary, but those words alone don't control what the agent buys. The payment policy sets the actual spending restrictions.
And this is where I still have a problem.
A $2 payment limit can stop a $5 transaction, but it won't stop the agent from making four useless $2 purchases.
It could follow every rule I approved and still waste most of my money.
I can see myself using Kite Passport for smaller tasks because I don't want to sit there approving payments all day. But I'd want to check what the agent actually paid for before trusting it with a bigger budget.
Keeping my money within a limit is good. Spending it on something worth buying is a different matter.
If your agent wasted $8 from a $10 budget without breaking any rules, would you give it another chance or start approving its payments yourself?
@KITE AI $KITE #KiteBBW