Application/Client Management

To provide programmatic access to the Aidbox API, you have to use a Client resource. The primary purpose of the Client resource is to facilitate authentication and access control within Aidbox.

In Aidbox, a Client represents an application or service that needs to interact with the FHIR server programmatically. Clients are using Basic and OAuth2 authentication flows to authenticate requests to Aidbox.

Unlike User resources, which represent human users, Clients typically represent:

  • Backend services

  • Mobile applications

  • Web applications

  • Integration engines

  • Other FHIR servers

Each Client has its own identity (client ID) and authentication mechanism (such as a client secret). This allows Aidbox to identify which application is making a request and apply appropriate access controls.

Client Access Control

Clients are subjects in the Aidbox access control system. This means:

  1. Each Client can be linked to specific AccessPolicy resources

  2. Different Clients can have different permissions

  3. API access is restricted based on the authenticated Client's permissions

This granular control allows you to limit what each application can do in your Aidbox instance. For example, a reporting application might only have read access to specific resources, while an EHR system might have broader read/write permissions.

Common Client Use Cases

API Integration Clients

Services that need to exchange data with Aidbox programmatically:

  • EHR systems

  • Analytics platforms

  • Health information exchanges

Single-Page Applications

JavaScript applications running in a browser:

  • Patient portals

  • Administrative dashboards

  • Clinical viewers

Mobile Applications

Native apps on mobile devices:

  • Patient apps

  • Provider apps

  • Care management tools

Backend Services

Server-side applications performing automated tasks:

  • Notification services

  • Data processing pipelines

  • Reporting engines

Security Considerations

When creating and managing Clients, consider these security best practices:

  • Assign the minimum necessary permissions

  • Use secure client secrets (high entropy, regularly rotated)

  • For public clients (like SPAs), use authorization_code with PKCE

  • Specify allowed redirect URIs for OAuth2 flows

  • Consider token lifetimes and refresh policies

Last updated

Was this helpful?