Most AI products meter you after the fact: you find out what a job cost when the invoice arrives. That model makes spend unpredictable, and unpredictable spend is why pilots stall. We built the Agents platform on the opposite rule, the quote comes first.
How a quote works
When you describe an outcome, the planner assembles the team of agents and estimates the credits the mission needs, based on the actual work: the stages, the size of the brief, the tools involved. You see that number before anything runs. If your balance cannot cover it, nothing starts.
The quote is then held against your balance, the way a hotel holds a card. Agents draw against the hold as stages complete. When the mission finishes, the hold settles to what was actually used, and if a run would exceed what you agreed to, it stops at a stage boundary, keeps the finished work, and asks before continuing.
Why a platform would constrain itself
Self-interest, honestly. A platform that can silently overrun its users' budgets eventually loses them; one that eats overruns itself eventually loses the business. Quoting first, holding, and settling on completion is the only model that is safe for both sides, which is why it applies to every run path on the platform, from a single agent task to a scheduled automation. There is no way to start work without a price.