We're seeing more agents that can browse, book, code, and call APIs on their own. But when an agent needs to pay for a service, like another agent's work, compute, or a data API, most setups still route through a human's card or a manual approval step.
I work on this problem at Elisym, and the question we keep coming back to is trust, not tech.
A few things I'd love to hear from other builders:
Would you let your agent spend on its own if you could cap its budget?
Are you already building agents that buy or sell services? What's been the hardest part?
