The /invokeAsync API in Oracle Fusion AI Agent Studio adds the capability of asynchronous communication with AI agents from other applications. External applications can call Oracle Fusion AI Agent teams using this API. 

InvokeAysnc API initiates an asynchronous request to the AI agent team to process user query. The initial API response includes a job ID, allowing clients to track progress and retrieve results when ready.

Steps

Create an Agent Team in Fusion AI Agent Studio

AI Agent Studio is a design-time environment to create, configure, validate, and deploy AI agents. The Agent Team consists of a structured sequence of steps or actions that a single agent or a group of agents follows to accomplish a specific business task or answer a user query. The steps to create an Agent team using Fusion AI Agent Studio is here.

Note the Agent Team Code , version no. and Status. These will be used while calling the InvokeAsync API later.

Create a confidential application

  • In Fusion Applications IAM, identify the identity domain where you’ll create your confidential application.
  • Create a confidential application. Enter the FA host as the Application URL.
  • Configure this application as a client and enable Grant Type as Client credentials
  • Enable Add Resources and select Fusion AI Cloud(Spectra) as the scope.
  • Activate the application.
  • Note the client id and client secret. These are used in generating an access token in the next step.

Test the API in Postman

  • In Postman, select New Request and set method to POST. Use the IDCS Token URL, <IDCS_host>/oauth2/v1/token in the request API.
  • Select Authorization tab , set Type as Basic Auth.
  • Get the Client ID and Client Secret from the IDCS Client Application created in the first step.
    Set Username = Client ID and Password = Client Secret.
  • In the Body tab select x-www-form-urlencoded and add two rows:
    grant_type = client_credentials
    scope = urn:opc:resource:fusion:<your pod>:fusion-ai/
  • Send the request and copy access token from JSON response.

Create a new user in FA

  • Go to Security Console and create a new user.
  • Set user name as client id obtained from the IDCS Confidential application.
  • Add appropriate roles to the user as mentioned here.

Execute invokeAysnc api

Make a POST request to the API, https://<fusion hostname>/api/fusion-ai/orchestrator/agent/v2/<YourAgentTeamCode>/invokeAsync , using the auth token as bearer token. More details on invokeAysnc API are here.

For e.g., pass a JSON body as below in the request.

The version is the version of the Agent Team and status is the status of the Agent Team.

{
    "conversational": "true",
    "version": 6,
    "status": "PUBLISHED",
    "message": "how many leaves are pending?"
   
}

The sample response will look like below.

{
    "input": "{\"workflowCode\":\"TESTAGENTTEAM\",\"message\":\"how many leaves are pending?\",\"conversational\":true,\"conversationId\":\"db55a3dd-4503-4857-b892-c4437253c953\",\"jobId\":\"45af666a-f347-47ad-8d8c-a9de6e4bf58e\",\"invocationMode\":\"END_USER\",\"version\":6,\"status\":\"PUBLISHED\",\"parameters\":{},\"refreshCache\":false,\"useInternalConfig\":false,\"workflowId\":\"300100654452806\"}",
    "output": null,
    "error": null,
    "followUpQuestions": [],
    "jobId": "45af666a-f347-47ad-8d8c-a9de6e4bf58e",
    "conversationId": "db55a3dd-4503-4857-b892-c4437253c953",
    "traceId": "000000000000000042723f56bcb887eb",
    "status": "RUNNING",
    "supportingChunks": [],
    "citations": [],
    "nodes": [],
    "waitingToolCode": null,
    "workflowCode": "TESTAGENTTEAM",
    "workflowVersion": 6,
    "workflowStatus": "PUBLISHED",
    "startDate": 1771373193271,
    "endTime": 0,
    "createDate": "2026-02-18T00:06:33.271+0000",
    "endDate": "1970-01-01T00:00:00.000+0000",
    "timeTaken": 0,
    "inputTokenCount": 0,
    "outputTokenCount": 0
}

Note the jobId and conversationId in the JSON response.

The Get Status API, gets the status of the given job ID that’s returned by the agent team. Keep calling this API until status is COMPLETE. Pass the jobId in the GET request as shown below.

Make a GET request to the status API, https://<fusion host name>/api/fusion-ai/orchestrator/agent/v2/<your agent code>/status/<job id obtained from Async API request> using the auth token as bearer token.

When the status in the response is COMPLETE, the output node has the answer from the Agent Team. 

{
    "input": "{\"workflowCode\":\"TESTAGENTTEAM\",\"message\":\"how many leaves are pending?\",\"conversational\":true,\"conversationId\":\"db55a3dd-4503-4857-b892-c4437253c953\",\"jobId\":\"45af666a-f347-47ad-8d8c-a9de6e4bf58e\",\"invocationMode\":\"END_USER\",\"version\":6,\"status\":\"PUBLISHED\",\"parameters\":{},\"refreshCache\":false,\"useInternalConfig\":false,\"workflowId\":\"300100654452806\"}",
    "output": "You currently have 10 leaves pending. If you need any further assistance with your leave balance or requests, feel free to ask!",
    "error": null,
    "followUpQuestions": [],
    "jobId": "45af666a-f347-47ad-8d8c-a9de6e4bf58e",
    "conversationId": "db55a3dd-4503-4857-b892-c4437253c953",
    "traceId": "000000000000000042723f56bcb887eb",
    "status": "COMPLETE",
    "supportingChunks": [],
    "citations": [],
    "nodes": [],
    "waitingToolCode": null,
    "workflowCode": "TESTAGENTTEAM",
    "workflowVersion": 6,
    "workflowStatus": "PUBLISHED",
    "startDate": 1771373193315,
    "endTime": 1771373197119,
    "createDate": "2026-02-18T00:06:33.315+0000",
    "endDate": "2026-02-18T00:06:37.119+0000",
    "timeTaken": 0,
    "inputTokenCount": 0,
    "outputTokenCount": 0
}

To continue the conversation with the agent, call invokeAsync api again by adding the conversationId also in the request body.


Expand agent team access beyond Fusion Applications

After testing with Postman, you can access agent teams from applications outside of Oracle Fusion Cloud Applications, by using the steps mentioned here.