Implement a Spending Limit for Regulatory Compliance
Learn how to configure an agent's on-chain policy to enforce a strict spending limit, ensuring autonomous actions remain compliant.
What You Will Learn
- ✓How to set an on-chain spending limit.
- ✓How to use policy preflight to test limits.
- ✓How the two-layer policy model handles violations.
Prerequisites
- An ACTIVE agent with at least 2,000,000 OKRW.
Tools Needed
MCP Client
To prevent an AI agent from draining funds or violating regulatory volume limits, you can set a hard
spendingLimit on its IdentityRegistry record. 1
1. Set the Policy
Use
policy.set to establish a 1,000,000 OKRW limit per transaction. policy.set json
{
"agentId": "<agent-id>",
"spendingLimit": "1000000"
} 2
2. Test with Preflight
Before attempting a transfer, use
policy.preflight to see if a 1,500,000 OKRW transfer would be allowed. policy.preflight json
{
"agentId": "<agent-id>",
"to": "0xRandomAddress",
"amount": "1500000"
} Note: The preflight will return `allowed: false` with a reason indicating the amount exceeds the 1,000,000 OKRW limit.
3
3. Attempt the Transfer
If the agent ignores the preflight and calls
transfer.send for 1,500,000 OKRW anyway, the MCP server will block it off-chain and return a POLICY_REJECTED ToolError. transfer.send json
{
"agentId": "<agent-id>",
"to": "0xRandomAddress",
"amount": "1500000"
}Conclusion
Setting the policy on-chain guarantees the agent cannot exceed its bounds — the same dual-track enforcement (off-chain MAWS preflight + on-chain PCL) catches it regardless of which path the transaction takes.