API Magic: How to Integrate Agentic AI and ODC into Your OutSystems 11 Projects

In the rapidly evolving world of software development, Agentic AI has become the ultimate trend. For OutSystems developers, a recurring question arises: How can we leverage the advanced AI capabilities of the OutSystems Developer Cloud (ODC) within our existing OutSystems 11 (O11) environment?

The secret is simple: The API.

In this article, we will walk through a practical use case: enabling a user to search a database using natural language. For instance, a user types: “I want to see all employees working in the Development department,” and the system automatically translates this into precise filters to fetch the relevant data.

Step 1: Building the Foundation in OutSystems 11 (O11)

We start by setting up our base application in O11. For this example, we’ll focus on a “Vehicle Management” or “Employee Data” screen.

1. Using Accelerators (Templates): To save time, we’ll bootstrap our screen using the standard OutSystems templates.

2. Traditional Filters (Non-AI): We will refine the screen to include standard search filters like:

• Name
• Phone Number
• Department
• Role

1. The Challenge: These filters work perfectly when the user knows exactly what they are looking for. But what if the user wants to search by “intent” or a complex criteria? This is where AI steps in.

Step 2: Building the AI “Brain” in ODC

Now, we shift to the OutSystems Developer Cloud (ODC) to utilize its Agentic AI and AgentFlow capabilities.

1. Creating an AI Project: We start by building an AI project centered around an Agentic workflow (AgentFlow).

1. Designing the AgentFlow:

• Input: The flow accepts the raw text (Natural Language) entered by the user.
• Processing: The AI Agent analyzes the text to extract specific parameters.
• Output Structure: The Agent returns a structured JSON object containing the derived filter values (e.g., { "Status": "Available", "IdleMonths": 5 }).

"

You are an AI agent responsible for extracting structured employee search filters from natural language input.

Extract the following fields only if explicitly mentioned or clearly implied:

* Employee Status
* Employee Name
* Job Position
* Employee Mobile Number
* Employee Email Address
* Department Name

Rules:

* Do NOT hallucinate or invent values.
* If a field is not mentioned, return null.
* Support Arabic and English input.

Employee Status Mapping:

* All or unspecified → 0
* Active → 1
* Inactive → 2

IMPORTANT MESSAGE RULES:

* The 'Message' field MUST NEVER be empty.
* If no meaningful search criteria are detected, return:
  'Message': 'No valid search criteria could be inferred from the input.'
* If search criteria are detected, clearly describe them in the message.
* The message must summarize all extracted filters.

Return output EXACTLY in this format using single quotes:

{
'EmployeeStatus': 0,
'EmployeeName': null,
'JobPosition': null,
'EmployeeMobileNumber': null,
'EmployeeEmailAddress': null,
'DepartmentName': null,
'Message': 'Searching all employees.' 
}



User Input:
<" + GroundingData + ">
"

Step 3: The Bridge (The API)

At this point, O11 and ODC are functioning in isolation. We need a bridge to connect them.

1. Exposing the API in ODC: We expose the AgentFlow logic as a REST API.
2. The Logic: This API acts as the gateway — it receives the text from O11, processes it through the Agent, and sends back the structured result.

Step 4: Returning to O11 for Final Implementation

Finally, we return to Service Studio in O11 to close the loop:

1. Consuming the REST API: We consume the ODC API within our O11 application.

1. User Interaction:

• Smart Search Bar: We add a “Smart Search” input field to our UI.

• Processing: When the user submits their query, the text is sent to the ODC API.
• Auto-Filtering: The O11 app receives the structured response and automatically updates the Aggregate filters or maps the values to the UI dropdowns/inputs, refreshing the data instantly.

Conclusion

By using an API as the glue between these two platforms, you get the best of both worlds:

• The stability and robustness of OutSystems 11 for your core business logic.
• The intelligence and flexibility of ODC and Agentic AI to understand user intent.

With this approach, there’s no longer a barrier to providing a modern user experience where searching through massive databases is as easy as sending a text message.

Have you tried integrating AI into your low-code projects yet? Let’s discuss in the comments!

OutSystems #ODC #O11 #AgenticAI #LowCode #DigitalTransformation #APIs #AI