Compare AI agent resource types
This table helps you compare the resource types that you can use in a managed connection to secure an AI agent. It highlights key differences in the credential type, security protocols, and security levels. Use this information to evaluate which approach best aligns with your org environment and requirements.
| Resource type |
Use case |
Credentials type | Access control | Connection protocol | Security level |
|---|---|---|---|---|---|
| Authorization server | Securing Homegrown: Fully custom AI agents to internal APIs | Policies and scopes | Authorization server configuration | Cross App Access | High |
| MCP server | Securing Homegrown: Fully custom AI agents to SaaS resources | OAuth tokens | Secure Token Service | OAuth | Medium |
| Application | Securing Homegrown: Fully custom AI agents to apps APIs that use OAuth consent flow on behalf of a user | OAuth tokens | Secure Token Service | OAuth | Medium |
| Secret | Securing Homegrown: Fully custom AI agents to SaaS resources that require an API key to access resources | API tokens | Okta Privileged Access (OPA) | AI token exchange | Low |
| Service account | Securing homegrown, fully custom AI agents to SaaS API resources that require basic authentication | Username and password | Okta Privileged Access (OPA) | AI token exchange | Low |