Deep Research Agent
Deep Research Agent researches a topic in depth. When you ask it to research something, it first prepares a plan with several angles on the topic, then searches for information on each one with the search tools you connect and, finally, writes an extensive, structured report that brings together everything it found.
Use it when you need a thorough study rather than a quick answer. Because it works step by step and consults many sources, it takes longer than a regular agent.
When to use it
- A market study of a sector or a product.
- A competitor analysis before launching a service.
- A report on a regulation, a technology or a trend.
- Background material to prepare a sales proposal.
Configuration
| Connection | What to connect |
|---|---|
| Model (required) | An AI model component. For long reports, a model with good reasoning ability works best. |
| Tools | The search tools it will use to research, for example Google Serper API, Brave Search API Tool or DuckDuckGo Search. Without search tools, the agent cannot consult sources. |
| Field | What it does |
|---|---|
| Agent Name | The agent’s name when another agent uses it as a tool or subagent. Default: DeepResearchAgent. |
| Agent Description | What the agent does. Another agent reads it to decide when to turn to it. |
| User Prompt | The template of the message the agent receives. Default: {input}. Each variable in curly braces becomes a field with that name. |
| Input | The topic or question to research, usually from Chat Input. |
| Target Language | The language of the final report. With Auto-detect from input (default) it answers in the language of the request. You can also choose English, Spanish, French, German, Italian, Portuguese, Chinese (Simplified), Japanese, Korean, Russian or Arabic. |
Advanced options
| Field | What it does |
|---|---|
| System Prompt | Extra instructions, for example the report’s audience or the aspects it should focus on. Empty by default. |
| Recursion Limit | The maximum number of internal steps in one answer. When it is reached, the agent stops. Default: 1000. |
| Stream | The report appears in the chat as it is generated. Turning it on shows Stream Thinking, which also shows the model’s reasoning. |
| Use memory | The agent remembers the conversation, so you can ask it to extend or change the previous research. Turning it on shows the Memory connection. |
| Include State in response | Adds the agent’s internal run data to the answer. |
Using it as an agent tool
You can connect this component to an agent’s Tools input. In the Tool tab, Tool Name and Tool Description are the name and explanation the agent uses to recognise the tool and decide when to call it: write a clear description of what it does.
Batch processing
With the Batch Mode switch on, the component runs once for each item of the list you connect to Batch Input and returns all the results together as a list. Parallel Executions sets how many items are processed at the same time (1 to 5). Skip if Empty Batch Input skips the component when the list arrives empty, and Skip Branch if Empty Output stops the following components when the result is empty. More in Key concepts.
Output
- Response: the final report as a message. Connect it to Chat Output or to a component that saves it as a document, such as Docx Writer or Pdf Writer.
- Agent: the configured agent, so another agent can delegate to it.
- Tool: the agent turned into a tool, for another agent’s Tools input.
Example
The travel agency Rumbo Sur wants to launch rural tourism getaways. It connects a web search tool to Tools, chooses Spanish in Target Language and writes in the chat: “Research demand for rural tourism in Spain: traveller profile, most searched regions, usual prices and recent trends.” The agent prepares its research plan, consults sources on each point and delivers a complete report, which the agency saves as a document with Pdf Writer.
Related components
- Deep Agent: for long tasks of any kind, not only research.
- Agent: for quick answers with tools.