Skip to main content
Connect your application to the Agent API using your existing backend language and HTTP client. Start with account access, then add the identity, contract, token, or payment flow your product needs.

​
1. Connect an account

Choose the flow matching your credentials: Use the exact host for that account. Send your application’s HTTPS URL as Referer on every Agent API call, including requests from a backend.

​
2. Connect your interface to the API flow

For a web or mobile product, let your backend coordinate the Agent API requests: Map each application user to the Neuro account and objects they may use. Keep account passwords, API secrets, signing-key passwords, and JWTs in protected backend storage. Return the fields your interface needs to display the outcome. For user-interactive account onboarding, the browser signup flow describes the supported form and session handoff.

​
3. Add your first capability

Each guide shows the requests, prerequisites, and result to check: Present the exact contract and role before signing, and the provider, amount, and currency before initiating a payment. Submit the action after the user confirms those details.

​
4. Track completion

Store each returned identifier with the corresponding application record and Neuron host. Some requests start work that completes later: an identity can await approval, a contract can await signatures, and a payment can remain pending. Use the retrieval operation named in the capability guide to update your interface. Apply bounded polling and show a pending state until the required result is observed. See asynchronous state. If a write times out, retrieve the known object or reconcile with the provider before submitting it again. An absent response does not prove the operation failed.

​
5. Verify the application flow

Check that your interface shows the intended account and object, handles expired sessions, and distinguishes pending work from success or failure. Exercise a rejected request and an unavailable service as well as the successful sequence. Your first integration is complete when a user can perform the chosen action, see its confirmed result, and return later to the same object.