#1176 OIC Fusion AI Agent Studio Adapter

Introduction 

This is a feature introduced in the 26.07.2 release. 

Kudos to my esteemed colleague Madhav P. and team for getting this out so quickly.

With this adapter, we can invoke a Fusion AI Studio Agent. Let's check out one of the agents on offer, before leveraging the adapter to do so.

Fusion AI Studio - Agents

I access these via the Workflows option in Fusion AI Studio. I run the OOTB, Employee Onboarding agent -

My prompt - please provide the person numbers, names, email ids, projected start dates and manager names of all pending employees whose projected start date is 2026-09-17 

The agent response - 

Thank you for sharing the pending worker details. Here's a summary of the employees to be onboarded with the projected start date of 2026-09-17:

1. Rahul Mehta | Email: rahul.mehta@acme.com | Manager: E10
2. Olivia Thompson | Email: olivia.thompson@acme.com | Manager: E10
3. FlorenceTest56 HumphreyTest56 | Email: 56jewelhumphrey@acme.com | Manager: E10
4. Ananya Iyer | Email: ananya.iyer@acme.com | Manager: E10
5. Michael Williams | Email: michael.williams@acme.com | Manager: E10

Please let me know if you need assistance with onboarding steps, document preparation, or any other details!

Let's try out the same, using the new adapter.

Create the Connection

Security is JWT based. The pre-requisites for this, e.g. confidential app setup etc. are documented in the adapter guide. Check it out here.

I add the required input, then test the connection -

I have based mine on a shared connection; now to the integration.

Create the Integration

This is an app driven integration with a REST Trigger.

Send message and Get task are two of the operations offered by the A2A protocol.

Send message does as it says - sends a single payload/instruction to the target agent. The operation returns immediately with either a direct message response or initial task metadata, e.g. task id, if the request requires background processing.

Get task - can be used for polling the status of a particular task.

We will use both of these in the integration.

Firstly, Send message to Employee Onboarding - 

I set role to user and map my request prompt to text.

For demo purposes, I add a WAIT of 20 secs.

Then I invoke the agent again, this time to GetTask. 

Source Field for mapping - 

Target field - 

Now all I need is to map the integration response; the source is - 

Let's run the integration - 

I provide the same request I used directly in Fusion AI Agent Studio.

The response - 

{

  "response" : "Thank you for sharing the pending worker details. Here's a summary of the employees to be onboarded with the projected start date of 2026-09-17:\n\n1. Rahul Mehta (Person Number: 7690), Email: rahul.mehta@acme.com, Manager: E10  \n2. Olivia Thompson (Person Number: 7691), Email: olivia.thompson@acme.com, Manager: E10  \n3. FlorenceTest56 HumphreyTest56 (Person Number: 7683), Email: 56jewelhumphrey@ACME.com, Manager: E10  \n4. Ananya Iyer (Person Number: 7693), Email: ananya.iyer@acme.com, Manager: E10  \n5. Michael Williams (Person Number: 7688), Email: michael.williams@acme.com, Manager: E10  \n\nLet me know if you need assistance with onboarding steps, documentation, or any other details!"

}

Summa Summarum

This adapter makes it so easy to interact with Fusion AI Agents. Those agents can also easily leverage OIC based MCP tools, making this a very compelling ecosystem for your AI agents.

Cerevisia Aestiva

Today's beer, or should I say ale, is from the UK. I've just spent a very exciting week there, working on Agentic AI projects with some of our top UK customers.

The ale comes from Sharps brewery, which is located in Cornwall. The name refers to a treacherous sandbank in the locality. Doom Bar has a taste, second to none; a compelling combination of hops and malts, that is almost fruity, but with a light bitter aftertaste. 

Try it out, next time you're in the UK; and, if you're UK based, try it out tonight!