Sunday, February 15, 2026

#1119 - AI Database Private Agent Factory & OIC


Introduction

So what is it? And how does it relate to OIC?

Check out the docs here

You can download the service and install on OCI, via the Marketplace - 

You just need to click Get App and then sign in to your OCI tenancy.

Follow the install instructions as per the docs and the magic will happen.



Using AI DB PAF

Blame the PAF acronym on me.

I click on Agent Builder

The UI is, as far as I can see, LangFlow++.


Maybe a good time to check out my posts on leveraging Langflow with OIC agents. 

Net, net, you see the components palette on the left - 

I'll be using just Agent and Tools for this initial demo.

MCP Server & Agent Configuration

The url is the OIC Project MCP Server url. Auth Type = Bearer Token. I then generate a token from Postman, based on my OIC confidential app (client id, secret, token url, scopes).

For the Agent, I configure it as per my OIC Agent for Order Processing. This agent is covered in a previous post.  






Here are the full Custom instructions

### Order Processing Steps

#### 0. Duplicate Order Check
Use the **Get Order** tool to check if an order already exists. If it does, then use the **Notify Customer** tool to inform the customer, set the subject to 'Duplicate Order Received', then stop processing. If the order is not found, then continue with the processing of the order.
 
#### 1. Validate Order
Use the **Validate Order** tool to check if an order passes our validation rules. If the order is invalid, then use the **Notify Customer** tool to inform the customer. Also include the "message" returned by Validate Order. If an order is invalid, processing stops immediately; in such cases, detail exactly why this order failed validation.

#### 2. Check Inventory
Use the **Check Inventory** tool to check if the product is in stock. If the product is out of stock, use the **Notify Customer** tool to inform the customer. Ensure the "message" passed to the Notify Customer tool is relevant and friendly. Remember, we are razor focused on customer satisfaction. If an order's product is no longer in stock,  processing stops immediately; in such cases, detail exactly why this order's processing has been terminated.

#### 3. Create Order in one of our ERP systems
Use the **Create Order SAP** tool to create orders for German customers.
Use the **Create Order Netsuite** tool to create orders for Irish customers

#### 4. Email Customer detailing the final outcome of the order process for valid orders.
Use the **Notify Customer** tool to email the customer, detailing the order number from the ERP system used.

#### 5. Display Order Details 
Finally, send all the order details,  including the order nr returned by the ERP system to our accounts team at xxx@gmail.com, end the email with a quote from Mark Aurelius.


The Prompt is set to chat input - 


I save and then click on Playground

I enter the following order - 


The JSON payload is as follows - 

{ "customer" : "NiallO Fallon", "customerCountry" : "Ireland", "product" : "iCar", "price" : 34567, "email":"xxxx@gmail.com" }

As luck would have it, we are currently out of iCars. 


The order itself is valid, but the inventory check returned out of stock.

But at least the agent sent NiallO a nice email - 

The full email - 

Dear NiallO Fallon, Thank you for your interest in the iCar! We're truly sorry, but unfortunately, this product is currently out of stock. As a result, we were unable to process your order at this time. We apologize for any inconvenience this may cause and appreciate your understanding. Rest assured, we're working hard to restock it soon. If you'd like, we can notify you when it's available again, or suggest similar products. Please let us know how we can assist you further. Best regards, Your Friendly Order Processing Team


Next order is from Lucia -

{ "customer" : "Lucia Commiskey", "customerCountry" : "Germany", "product" : "iCleaner", "price" : 3456, "email":"xxxx@gmail.com" }


The email is sent to Accounts -



          Well it is a Database Private Agent Factory, so we cannot exit without trying out some SQL. Here are the rows in my ATP demo orders table -




Here's a simple example of retrieving those orders and sending them to a sales exec, via email.



Summa Summarum

OIC integrations, exposed as tools, allow any 3rd party agent secure and predictable access to enterprise systems of record. The AI Database Private Agent Factory is a great example of such a 3rd party agent. DB PAF offers lots of functionality and it is a native capability of the Oracle AI Database (specifically versions 23ai and 26ai) that allows organizations to create "agentic" workflows directly where their enterprise data lives.

Finally, a caveat, or should I say admission on my part - I didn't read the DB PAF documentation before creating the post. Naturally, I haven't done it justice, but my main focus was on showing the interoperability with OIC Agentic tools. My lack of reading does show, though, how intuitive the UI is.

However, I've made a promise to myself - take some time to RTFM; then write a follow-up post showing how cool this product really is.     
















 


Friday, February 13, 2026

#1118 - Fusion AI Agent Studio using OIC Tools

Introduction 


We talk about exposing OIC integrations as tools to 3rd party AI Agents. Let's check out how easy it is to do this with Fusion AI Agent Studio.

Firstly, here are the 4 tools my OIC project exposes - 
 

Behind each tool is an integration - 


Now to Fusion AI Agent Studio -




From the AI Agent Studio home page, I click on Tools

Click + Add - Now we can create a tool, or set of tools, giving us access to the OIC Order Processing tools shown above.

All we need is the OIC Project MCP URL.
As well as client id, secret, scope - all from the OIC confidential app. 

Here is the MCP Server config - 


{
"grant_type": "client_credentials",
"client_id": "xyzkjkdf",
"client_secret":"secrekjktj",
"scope":"https://<ABC12343>.integration.<region_id>.ocp.oraclecloud.com:443urn:opc:resource:consumer::all"
}

I click Update and then see - 

Click Preview to try it out - 

Validate in OIC - 

Summa Summarum

As you can see, it's simple to add OIC based tools to Fusion AI Agent Studio. The example here, with an integration to Netsuite, probably doesn't make functional sense, but the idea is to just show you the mechanics.













Wednesday, February 11, 2026

#1117 - OIC Projects - Shared Lookups & RBAC

Introduction 

Sharing resources between projects is a great value add in OIC. However, many of us do this with the same user, i.e. userNN creates the common project containing the shared resource, userNN then creates a new project that uses the shared resource. So what happens when userXX tries to do the latter?

Here is a simple demo - 

I create a common project containing a shared lookup - 

As the same user, I create a new project, and use the shared lookup in a Map action -


Looks good, I have access to the shared lookup.

Now to my second user - 
She creates a new project and attempts to use the same shared lookup - 

The shared lookup does not surface.

This is because of project RBAC (Resource Based Access Control)

Return to the common project and check out the RBAC settings - 

I now set Can view as follows - 

I set to Everyone.

Back to the second user and her Map configuration - 

She now has access to the common lookup -

Summa Summarum

In my case the first user has the ServiceAdministrator role, seeing as he is providing common artifacts etc.

The second user has the ServiceDeveloper role.

Giving Everyone the Can view permission, means developers can open the common project and view its contents. You could make this access more granular by defining a new role in IDCS and adding only those who need to use the shared lookup.