Skip to main content

Deploy and Query an AI Agent

This page shows how to deploy an AI agent to WSO2 Cloud - Integration Platform and query it after deployment. Use it when you want to test retrieval-augmented generation behavior from the built-in chat UI or the service API.

Prerequisites
  • WSO2 Integrator installed on your machine.
  • A working AI agent. You can follow the IT helpdesk chatbot tutorial.
  • API keys for any model providers used by the agent, for example OpenAI. Do not commit keys to source control.
  • Access to a WSO2 Cloud account with permission to create projects and services.

Overview​

You will:

  1. Create an AI agent.
  2. Deploy the agent using the WSO2 deployment guides.
  3. Configure required environment variables and secrets.
  4. Test the agent using the built-in chat UI or the service API.

Step 1: Create an AI agent​

Follow the tutorials on creating an AI Agent or import your own AI agent in WSO2 Cloud.

Step 2: Deploy the agent​

Follow the WSO2 deployment guides for step-by-step instructions covering publishing, granting repository access, and deploying from the Cloud editor or from your IDE.

Deploy the agent

Step 3: Configure environment variables and secrets​

  1. After the initial build, open the integration Overview page in WSO2 Cloud.
  2. Click Configure and add API keys and any required secrets.
  3. Apply the configuration and wait for the redeploy to complete.

Configure secrets and environment variables

Step 4: Test and query the agent​

There are two common ways to query the deployed agent:

  • Use the built-in chat UI available on the integration overview page.
  • Call the agent HTTP endpoint using the OpenAPI console or curl.

Example queries to exercise RAG and tools:

  1. "What is the leave policy for new hires?"
  2. "What is the annual performance review?"

Chat with the agent

Troubleshooting and verification​

  • If builds fail, open the build logs from the integration overview and inspect the failure reason.
  • Use the OpenAPI console (Test on the development card) to call endpoints directly and verify responses.
  • Check ingestion logs to confirm that FAQ documents were indexed into the vector store.

What's next​