How Do AI Agents Make Decisions?
Agents decide by combining model reasoning with tool access: what to do next, which tool to call — and whether to spend.
The decision loop
The model proposes the next action from context.
The runtime executes tools and returns results.
The loop repeats until the goal is met.
Where spending decisions happen
The model decides to call a tool; the tool may spend.
The model's proposal is a request — the gate is what enforces it.
Why enforcement can't be in the prompt
Prompts can be injected or ignored; rules can't.
A deterministic gate on the money path is the reliable control.
FAQ
Do agents decide to spend?
They decide to call tools — spending is a side effect the gate governs.
Can I trust the model's spending judgment?
Not alone — documented runaways show why the money path needs rules.
Related
Stop the next $12,400 night.
One API call (or MCP tool) in front of every agent transaction — APPROVED, BLOCKED, or FLAGGED, deterministic, ~5 ms, fully logged.
See plans — from $99/mo Try a live check