What you control
When an agent can spend, the important question is not just “can it pay?” It is “can it pay within the rules I set?” You should control:- which wallet funds the agent;
- which networks and assets it can use;
- which sellers, agents, or domains it can pay;
- how much it can spend per request, task, route, or time window;
- whether larger payments require approval;
- when the agent should stop, ask for help, or choose a cheaper option;
- how payments connect to task logs and delivered results.
Set up your agent
1
Create a dedicated wallet
Use a wallet for the agent or application, not your main wallet. Keep automated signing separate from personal funds.
2
Deposit collateral
Deposit ETH or a supported into 4Mica on the network your target sellers accept.
3
Register the client scheme
Configure the 4Mica client for each network the agent may use.
4
Define spending policy
Set limits for amount, seller, route, network, task, time window, and approval requirements.
5
Make paid requests
Use a wrapped HTTP client so the agent can handle paid resources while Your policy decides which payments to sign.
6
Audit the task
Record each payment with the seller, route, amount, reason, guarantee ID, request ID, and task result.
Make a Paid Request
Before the agent pays
Your agent should inspect the seller’s payment requirements and decide:- is this seller or domain allowed?
- is the price within budget?
- is the network and asset allowed?
- does the payment match the task?
- does the agent need human approval first?
- should the agent continue, stop, or choose another seller?
After the agent pays
Every paid request should produce a record that explains what happened:- which agent paid;
- which seller received the payment;
- how much was authorized;
- why the payment happened;
- which task or request it belongs to;
- what result came back;
- whether the guarantee or settlement obligation is still open.