we need o3-deep-research support in Responses APIs

denis.wen 0 Reputation points
2025-07-03T11:51:24.46+00:00

Currently, o3-deep-research is only supported for use with Azure Agents, and the process to call it is quite complicated. It does not support direct invocation via API key, and the code implementation is excessively complex, requiring the creation of agents and other operations, which makes integration inconvenient.

https://learn.microsoft.com/en-us/azure/ai-foundry/agents/how-to/tools/deep-research

Azure AI services
Azure AI services
A group of Azure services, SDKs, and APIs designed to make apps more intelligent, engaging, and discoverable.
3,685 questions
0 comments No comments
{count} votes

2 answers

Sort by: Most helpful
  1. Martin Egli 435 Reputation points
    2025-07-03T11:58:20.97+00:00

    Hi Denis,

    Thanks for highlighting this. Here's a concise summary and suggestion regarding the o3-deep-research tool support in API integrations:

    Issue Summary:

    The o3-deep-research tool is only available via Azure Agents.

    There is no direct API key-based access, making integration difficult.

    Current implementation requires:

    Agent creation and configuration

      Tool binding within Azure AI Studio
      
         Indirect invocation through agent orchestration
         
    

    This significantly raises the barrier for using the tool in lightweight or external applications via API.

    Suggested Improvement:

    To improve adoption and flexibility, Microsoft should consider:

    Expose o3-deep-research via direct REST API endpoint, authenticated with Azure OpenAI or Azure Foundry keys (similar to /chat/completions).

    Simplify developer onboarding by:

    Providing an SDK wrapper (e.g. in Python or JS)

      Offering CLI or cURL-based examples
      
         Allowing execution outside the agent flow (stateless)
         
         **Decouple agent logic from tools**, allowing developers to:
         
            Call tools like `deep-research` directly
            
               Use them with external orchestrators or apps
               
    

    Current Workaround (if staying in Azure AI Foundry):

    If you must use o3-deep-research today, you need to:

    Create an Azure AI Agent

    Attach the o3-deep-research tool

    Send user prompts via the agent conversation endpoint

    Wait for the agent to use the tool as part of its plan

    Docs: Deep Research Tool – Azure AI Agents

    Conclusion:

    The o3-deep-research capability is powerful, but current restrictions reduce its accessibility. Exposing it via direct API would dramatically improve developer experience and integration potential — especially for non-agent use cases.

    Let me know if you’d like help drafting a feature request or GitHub issue to Microsoft.

    Best regards,
    Martin


  2. Martin Egli 435 Reputation points
    2025-07-03T12:19:54.3666667+00:00

    Hi Denis,

    Thank you for your follow-up.

    As of now, Microsoft has not published an official timeline for exposing o3-deep-research via direct REST API outside of the Azure Agents framework. Currently:

    • The feature remains exclusive to the Azure AI Agent architecture.

    No roadmap item or public preview has been announced regarding API-key-based standalone access.

    However, based on the documentation and feedback loops from Azure AI Foundry:

    Microsoft is actively collecting user feedback to guide priorities.

    The modularization and tool decoupling from agents have been mentioned in internal discussions as a long-term objective.

    Recommendation:

    To help accelerate this:

    I recommend submitting a formal feature request via the Azure Feedback Portal or GitHub (if applicable).

    I’d be happy to help you draft that request or escalate through Microsoft support channels if you're working under a CSP or enterprise agreement.

    Note: I’m not a Microsoft employee - just a developer/architect who shares your interest in easier integration of o3-deep-research outside the agent ecosystem.

    Let me know if you'd like a ready-made draft for a feedback submission.

    Best regards, Martin

    0 comments No comments

Your answer

Answers can be marked as Accepted Answers by the question author, which helps users to know the answer solved the author's problem.