Creating a Deep Researcher agent

Hitachi iQ Studio User Guide
Version
1.0.x
Audience
anonymous
Part Number
MK-26HIQS001-00
ft:lastEdition
2026-03-31

Create a deep researcher agent that performs workflow‑based research and generates detailed reports using your selected knowledge sources.

  • Make sure you have administrator or agent builder permission.
  • Open knowledge base and check that each required collection displayed a green checkmark.
  • Verify that the necessary MCP servers are registered.
  • You have prepared your system prompt (role definition) and know which model to use.
  1. On the iQ Studio home page, go to AI Agents and click Create.
  2. On the Create an AI Agent page, select Deep Researcher agent type.

    The following table describes the additional information required to create a new agent.

    Field Description
    Type Select the Deep Researcher agent type.
    Import Agent (.json): Upload a JSON file that defining the deep researcher agent’s configuration, including its roles, tools, workflows, and permissions. After you upload the file, the system reads the configuration and installs the agent.

    Note: Use this option only if you have previously exported an agent as a JSON file.
    Name Human-readable agent name.
    Description Short summary of the agent’s purpose.
    Avatar Optional image representation.
    Avatar dropzone You can use a custom image. The supported file types are JPG, JPEG, JFIF, PJPEG, PJP, PNG, SVG, and the maximum file size must be under 1 MB.
    MCP prompt Define the agent’s behavior and tone.
    Allow Clarification Toggle to allow the researcher to ask clarifying questions before research.
    Advanced options* Optional settings used to control agent limits and behavior.
    Max Structured Output Retries Set the maximum number of retries for structured output generation.
    Max React Tool Calls Set the maximum number of tool calls during research.
    Max Concurrent Research Units Set the maximum number of research units that can run at the same time.
    Max Researcher Iterations Set the maximum number of iterations the researcher can perform.
    Summarization Model Select the summarization model.
    Summarization Model Max Tokens Specify the maximum token limit for the summarization model.
    Research Model Select the model used for conducting research.
    Research Model Max Tokens Specify the maximum token limit for the research model.
    Compression Model Select the model used for compressing research data.
    Compression Model Max Tokens Specify the maximum token limit for the compression model.
    Final Report Model Select the model used for generating the final report.
    Final Report Model Max Tokens Specify the maximum token limit for the final report model.
    Note*: Leave advanced options at default unless you need custom control policies.
  3. Click Next.
  4. On the Equip page of the Build a Deep Researcher AI agent wizard, add one or more agents that you created and then click Next.
  5. On the Configure Tool page, enter tavily as the search engine and enter the Tavily API key.
  6. Set the permissions for the tool:
    • Always Allow: Grants permission to run without asking for confirmation each time.
    • Ask Before Running: Requires user confirmation before the tool runs.
  7. Click Apply.
  8. Click Next and then click Done.
  9. Confirm that the new deep researcher agent appears in the AI Agents list on the AI Agents page.