Agent Spend Approval Policy Template
Every agent that can spend needs an approval policy: what's auto-approved, what waits for a human, and who's on call. This template gets you there.
1. Define thresholds
Auto-approve: transactions under [AMOUNT] that match the merchant allowlist.
Require approval: transactions over [AMOUNT], new merchants, out-of-hours spend.
Always block: banned merchants, over hard caps, over velocity limits.
2. Define the queue
Who reviews FLAGGED transactions: [OWNER / TEAM].
Response time target: [e.g. 15 minutes during business hours].
Escalation: if unreviewed after [TIME], page [ESCALATION PATH].
3. Enforce it
In sipi.bot: set approval thresholds per agent, merchant allowlists, and hard caps. FLAGGED transactions route to the queue automatically.
Example rules to set
{
"rules": [
{"type": "approval_threshold", "over": 500, "action": "flag"},
{"type": "merchant_allowlist", "allowed": ["openai.com", "anthropic.com"]},
{"type": "per_transaction_cap", "max": 5000, "action": "block"}
]
}Thresholds become rules; rules become decisions.
FAQ
What's a good default threshold?
Start with your largest 'no review needed' purchase — typically $100–$500 for dev teams. Adjust with data.
Who should own the queue?
The person who owns the budget: finance ops, a platform lead, or the team lead.
What happens to flagged spend overnight?
It stays in the queue until someone reviews — that's the point of approval.
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