How do I enable research mode in Brave's Answers API?

Definition: Brave's Answers API produces answers grounded in web search. As of 17 October 2026, research mode enables iterative searches through enable_research: true and requires stream: true.

Key facts

• Brave's Answers documentation describes research as sequential searches in which the model refines its strategy; response times can extend to minutes.
• The Answers API reference defines enable_research as a boolean that defaults to false.
• Brave's Answers integration guide documents progress events and a final answer containing citations in research mode.

How is research mode enabled in a request?

The Answers documentation places the research option in extra_body when using the OpenAI Python SDK. The client uses a Brave Search API key and Brave's base URL. The following example follows that documented configuration and reads the streamed content:

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_BRAVE_SEARCH_API_KEY",
    base_url="https://api.search.brave.com/res/v1",
)

stream = client.chat.completions.create(
    model="brave",
    messages=[{"role": "user", "content": "History of quantum mechanics"}],
    stream=True,
    extra_body={"enable_research": True},
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

In a direct JSON request, enable_research belongs in the request body. Brave's integration guide shows that form with stream set to true.

What changes in the answer workflow?

Brave's Answers documentation identifies single search as the default. Research mode performs sequential searches and processes a larger context, increasing cost. Brave recommends it for background tasks where thoroughness matters more than speed.

The integration guide describes a stream containing tagged messages. Its <progress> tag carries statistics such as iterations, queries and elapsed time. The <answer> tag contains the final synthesized answer; intermediate drafts are dropped. The <usage> tag at the end contains cost and token data. An application therefore needs to distinguish progress information from the final answer when displaying the response.

Which settings bound the research work?

The API reference exposes several research budgets. research_maximum_number_of_queries defaults to 20, research_maximum_number_of_iterations to 4, and research_maximum_number_of_seconds to 180. These respectively bound total searches, iterations and research time.

research_maximum_number_of_results_per_query controls the maximum results considered per search. Its default is 30, with an accepted range from 1 to 60. research_allow_thinking is accepted for compatibility but ignored: the engine runs research with extended thinking off.

Related terms

• Brave Answers API
• How do I read streamed citations from Brave's Answers API?
• Brave Search API
• Agentic search

FAQ

Should you also enable enable_citations?

Brave's API reference marks enable_citations as unsupported in research mode. You can use research mode's built-in answer citations, described in the integration guide, without enabling that separate option.

Can you use research mode with stream: false?

You must set stream: true. The same API reference explicitly requires streaming for enable_research.

Sources

1. Brave, "Answers", accessed 17 October 2026
2. Brave, "Answers API reference", accessed 17 October 2026
3. Brave, "Answers integration guide (skills/answers/SKILL.md)", accessed 17 October 2026

---

Check how a page reads to answer engines. Open the GEO score analyzer