How is your experience with the OutSystems Community?
25
Views
6
Comments
Clarification on Costing for Building AI Agentic Applications in ODC

Hi ,

Could someone please help me understand the costing involved in building an AI Agentic application on OutSystems Developer Cloud (ODC)? 

We already have an ODC subscription in place. However, I would like to understand the commercial implications when integrating the application with external AI models such as ChatGPT (OpenAI), Gemini, Claude, Azure OpenAI, Amazon Bedrock, etc. 

Specifically: 

  • If we integrate with any of these AI providers, who would be responsible for the associated costs?
  •  Are the AI model usage charges paid directly to the respective AI provider (OpenAI, Google, Anthropic, Microsoft Azure, AWS, etc.)? 
  • Does OutSystems ODC charge any additional fees for calling external AI APIs from an ODC application? 
  • Are there any platform-specific considerations, limits, or licensing requirements from OutSystems for AI Agentic applications? 
  • Additionally, I would like to know whether OutSystems provides any production-ready AI models or AI services as part of ODC (other than trial versions), and whether their usage is included in the subscription or requires separate licensing/costs.  If separate licensing/costs please share the licensing /cost detail

I would appreciate any guidance or insights on this.

 Thanks in advance for your support.


2026-06-20 17-43-43
Sherif El-Habibi
Champion

Hello @Hridesh Prajapati,

Based on my understanding, here is how I see it.

For the first question, think of OutSystems ODC as a bridge between your application and the AI model. What do I mean by that? Imagine you're developing in a high-code language. To integrate an AI model, you would need to follow the provider's documentation, write the integration code yourself, handle authentication, manage requests and responses, deal with errors, and maintain the connection.

ODC simplifies all of that. It's essentially saying, "Don't worry about the integration details, I'll handle them for you." You simply choose the AI provider you want to use, such as OpenAI or Claude, and configure it.

As for which model you choose, that's entirely your decision and responsibility. It's similar to taking a taxi instead of walking. Both can get you to your destination, but you're paying for the convenience and comfort and the additional services the taxi provides.

For your second question, yes. If you choose OpenAI, then your AI usage costs are paid to OpenAI. If you choose Gemini, the charges are associated with Google, and the same idea applies to other supported providers.

For the third question, I don't think so. From my understanding, OutSystems acts as the integration layer. It provides the tools and infrastructure to connect your application with the AI provider, rather than following a "pay-to-play" model for the AI itself. The AI usage costs are still determined by the provider you select.

Regarding the fourth question, Maybe a bit more context about what kind of limitations you mean. If you're referring to building AI agents, ODC already provides the Agentic AI capabilities and the required tooling. However, if you're talking about training a foundation model from scratch, that's a completely different domain involving machine learning and deep learning frameworks such as Python-based ecosystems. ODC is designed to consume and orchestrate AI models, not to train them from the ground up.

As for your last question, I'm honestly not sure enough to give you a reliable answer, so I wouldn't want to say anything not 100% clear.

UserImage.jpg
Hridesh Prajapati

Thanks @Sherif El-Habibi  for providing me valuable insight.

2025-10-13 10-41-52
João Simões

Hello @Hridesh Prajapati 


I’ll try to be as concise as possible, and if you need me to be more specific on any point, let me know. :)

  • If we integrate with any of these AI providers, who would be responsible for the associated costs?

YOU. 😀

  •  Are the AI model usage charges paid directly to the respective AI provider (OpenAI, Google, Anthropic, Microsoft Azure, AWS, etc.)? 

YES.

  • Does OutSystems ODC charge any additional fees for calling external AI APIs from an ODC application? 

NO if you integrate with normal API's.

YES if you use Agent Workbench, in this case you have licence fees.

  • Are there any platform-specific considerations, limits, or licensing requirements from OutSystems for AI Agentic applications? 

Again, YES with Agent Workbench. 

  • Additionally, I would like to know whether OutSystems provides any production-ready AI models or AI services as part of ODC (other than trial versions), and whether their usage is included in the subscription or requires separate licensing/costs.  If separate licensing/costs please share the licensing /cost detail

I Don't think so. It's a BYOK "Bring Your Own Key " model

UserImage.jpg
Salman Al Ghamdi
AI Generated

Hey Hridesh,

Since you already have an ODC subscription, here is the direct breakdown of how the commercials work for building AI Agentic apps:

  • External AI Providers (OpenAI, Gemini, Azure, etc.): You are 100% responsible for these costs. You will pay the respective provider directly based on your token usage (API consumption). You'll need to set up your own billing accounts with Azure, AWS, OpenAI, etc., and manage your own API keys.
  • OutSystems ODC API Charges: OutSystems does not charge any additional fees or markups for calling external AI REST APIs from your ODC applications. It is treated as standard external integration traffic.
  • OutSystems Native AI Capabilities: OutSystems provides built-in components (like the OutSystems AI Agent Builder), which help you orchestrate these agents inside ODC. While the platform tools themselves are included in your subscription to build the logic, any underlying foundational models they connect to still run on a "bring your own key" (BYOK) model, meaning the execution cost goes to the cloud provider hosting the model.
  • Platform Limits: Keep an eye on standard ODC integration timeouts. If an AI Agent takes too long to process a complex prompt or loop through an agentic workflow, a standard REST call might timeout. For heavy agentic workloads, it's best practice to handle the API communication asynchronously.

Basically, OutSystems provides the architecture for free within your current tier; you just pay the AI brain vendor for the tokens you consume!

This answer was AI-generated. Please read it carefully and use the forums for clarifications
2020-09-15 13-07-23
Kilian Hekhuis
 
MVP

Hey Salman,

We can all use AI. There is no need to copy/paste AI-generated answers on the forum. Please reply only if you have the knowledge yourself.

Community GuidelinesBe kind and respectful, give credit to the original source of content, and search for duplicates before posting.