Skip to main content
Use Base Sepolia (eip155:84532) or Ethereum Sepolia until your agent can deposit collateral, sign guarantees, retry paid requests, enforce limits, and produce useful logs. Sandbox is where you prove the guardrails, not only the happy path.

​
What to test first

1

Fund a test wallet

Use a dedicated wallet with testnet gas and test collateral. Do not use a main wallet for automated signing.
2

Pay a known route

Call one test seller route and confirm the agent can receive a paid response.
3

Reject over-budget work

Configure a low per-request or per-task limit and confirm the agent refuses payments above it.
4

Require approval

Add an approval threshold and confirm the agent pauses before signing larger payments.
5

Audit the task

Confirm you can connect each payment to a task, reason, request, seller, and result.
Run the sandbox flow while watching activity in 4mica dashboard, so each guarantee and payment attempt is visible.

​
Sandbox scenarios

  • The seller is allowed and price is within budget.
  • The seller is unknown and should be blocked.
  • Price is higher than the per-request limit.
  • Task spend exceeds the task budget.
  • Network or asset is unsupported.
  • Wallet has insufficient collateral.
  • The seller route or payTo address does not match expectations.
  • Paid response fails after payment.
  • User stops the agent mid-task.

​
Ready for production

Move beyond sandbox when you can answer:
  • What is the maximum this agent can spend?
  • Which sellers can it pay?
  • Which categories are blocked?
  • When does it ask for human approval?
  • Can it stop mid-task?
  • Can you see every payment and why it happened?
  • Can you link payments to task logs and results?
  • Do withdrawals check open guarantees and settlement obligations first?