Oracle Cloud Infrastructure Documentation

Managing API Endpoint Calling Tools in Generative AI Agents

An API endpoint calling tool is a custom tool in Generative AI Agents that enables you to securely integrate and call OCI APIs and your own REST APIs. Only HTTPS (secured) connections are permitted by Generative AI Agents.

By using API endpoint calling tools, you can make an agent a powerful extension of your existing APIs and business logic without custom integrations or changes to your backend systems. Through an uploaded OpenAPI schema that describes the structure and behavior of the API operations, Generative AI Agents can dynamically select and invoke the appropriate API endpoint and method during a chat in natural language.

The APIs can be hosted in public or private networks. All outgoing public or private HTTP traffic is routed through a subnet in your OCI virtual cloud network (VCN). The API operations can be requested with authentication or without using an authentication method, and then executed within your network's boundary and security rules.

By adding API endpoint calling tools to an agent in Generative AI Agents, the agent can call existing APIs to perform tasks such as retrieving customer data, updating order statuses, triggering workflows, or checking inventory. For example, a support agent could resolve a ticket by looking up a user account and initiating a refund, all through API calls.

Before You Begin

Before you create an API endpoint calling tool, ensure that you review the API Endpoint Calling Tool Guidelines and complete any tasks that are required.

Manage API Endpoint Calling Tools

If you have manage permissions to genai-agent-family, you can create and manage API endpoint calling tools in Generative AI Agents. See the following links:

See the example in API Endpoint Calling with SDKs in Generative AI Agents to learn how to create and invoke an API endpoint calling tool using OCI Java SDK.