Clear policy
Rules a machine can actually enforce.
Machine-readable policies define your agent’s authority and permit automatic approvals within their scope. Natural-language instructions alone do not enforce permissions.
Proposed transaction
Basex402 spending limit
25USDC
The limit you set
- Covers
- Paid APIs for Lisbon trip research
- Total
- Kept within your limit
- Network fees
- Review before signing
Telling an agent to “only spend a little” is a hope, not a permission. The model can misread it, drift from it, or be talked out of it, and nothing in the system stops the transaction. A policy is different: it is machine-readable, evaluated before a request ever reaches your approval, and it either permits the action or it does not.
Inside a policy’s scope, approvals can happen without interrupting you. Outside it, the request comes back to you — which is the point. The words you write are the interface; the rules they compile to are the enforcement.
Planned policy types
x402 spending controls
COMING SOONBounded spending an agent can settle against directly, scoped by amount, asset and counterparty.
Policies defined in natural language
COMING SOONAuthor a policy in plain words and have it compiled into the machine-readable rules that do the enforcing.
Policy controls are in development and are not yet enforcing permissions in the beta.
Use Superfluid Wallet with your agent.
Get started