Answers

Answers — AI Grounding

Requires API Key: Get one at https://api.search.brave.com

Plan: Included in the Answers plan. See https://api-dashboard.search.brave.com/app/subscriptions/subscribe

When to Use

This endpoint (/res/v1/chat/completions) supports two modes:

• Single-search (default): Fast AI-grounded answer from a single search. Supports enable_citations.
• Research (enable_research=true): Multi-iteration deep research with progress events and synthesized cited answer.

Quick Start (cURL)

Blocking (Single-Search)

bashCopy code
curl -X POST "https://api.search.brave.com/res/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "X-Subscription-Token: ${BRAVE_SEARCH_API_KEY}" \
  -d '{
    "messages": [{"role": "user", "content": "How does the James Webb Space Telescope work?"}],
    "stream": false
  }'

Streaming with Citations (Single-Search)

bashCopy code
curl -X POST "https://api.search.brave.com/res/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "X-Subscription-Token: ${BRAVE_SEARCH_API_KEY}" \
  -d '{
    "messages": [{"role": "user", "content": "What are recent breakthroughs in fusion energy?"}],
    "stream": true,
    "enable_citations": true
  }'

Research Mode

bashCopy code
curl -X POST "https://api.search.brave.com/res/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "X-Subscription-Token: ${BRAVE_SEARCH_API_KEY}" \
  -d '{
    "messages": [{"role": "user", "content": "Compare quantum computing approaches"}],
    "stream": true,
    "enable_research": true,
    "research_maximum_number_of_iterations": 3,
    "research_maximum_number_of_seconds": 120
  }'

Endpoint

httpCopy code
POST https://api.search.brave.com/res/v1/chat/completions

Authentication: X-Subscription-Token: <API_KEY> header (or Authorization: Bearer <API_KEY>)

SDK Compatible: Works with OpenAI SDK via base_url="https://api.search.brave.com/res/v1"

Two Modes

Parameters

Standard Parameters

Research Parameters

Constraints (IMPORTANT)

OpenAI SDK Usage

Blocking (Single-Search)

pythonCopy code
from openai import OpenAI

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

response = client.chat.completions.create(
    model="brave",
    messages=[{"role": "user", "content": "How does the James Webb Space Telescope work?"}],
    stream=False,
)
print(response.choices[0].message.content)

Streaming with Citations (Single-Search)

pythonCopy code
from openai import OpenAI

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

stream = client.chat.completions.create(
    model="brave",
    messages=[{"role": "user", "content": "What are the current trends in renewable energy?"}],
    stream=True,
    extra_body={"enable_citations": True}
)

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

Research Mode

pythonCopy code
from openai import AsyncOpenAI

client = AsyncOpenAI(
    base_url="https://api.search.brave.com/res/v1",
    api_key="your-brave-api-key",
)

stream = await client.chat.completions.create(
    model="brave",
    messages=[{"role": "user", "content": "Compare quantum computing approaches"}],
    stream=True,
    extra_body={
        "enable_research": True,
        "research_maximum_number_of_iterations": 3,
        "research_maximum_number_of_seconds": 120
    }
)

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

Response Format

Blocking Response (stream=false, single-search only)

Standard OpenAI-compatible JSON:

jsonCopy code
{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "choices": [{"message": {"role": "assistant", "content": "The James Webb Space Telescope works by..."}, "index": 0, "finish_reason": "stop"}],
  "usage": {"prompt_tokens": 10, "completion_tokens": 50, "total_tokens": 60}
}

Streaming Response

SSE response with OpenAI-compatible chunks:

textCopy code
data: {"id":"chatcmpl-...","object":"chat.completion.chunk","choices":[{"delta":{"content":"Based on"},"index":0}]}

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","choices":[{"delta":{"content":" recent research"},"index":0}]}

data: [DONE]

Streaming Tags by Mode

Single-Search (with enable_citations=true)

Research Mode

Usage Tag Format

The <usage> tag contains JSON-stringified cost and token data:

textCopy code
<usage>{"X-Request-Requests":1,"X-Request-Queries":8,"X-Request-Tokens-In":15000,"X-Request-Tokens-Out":2000,"X-Request-Requests-Cost":0.005,"X-Request-Queries-Cost":0.032,"X-Request-Tokens-In-Cost":0.075,"X-Request-Tokens-Out-Cost":0.01,"X-Request-Total-Cost":0.122}</usage>

Use Cases

• Chat interface integration: Drop-in OpenAI SDK replacement with web-grounded answers. Set base_url="https://api.search.brave.com/res/v1".
• Deep research / comprehensive topic research: Use research mode (enable_research=true) for complex questions needing multi-source synthesis (e.g., "Compare approaches to nuclear fusion").
• OpenAI SDK drop-in: Same SDK, same streaming format — just change base_url and api_key. Works with both sync and async clients.
• Cited answers: Enable enable_citations=true in single-search mode for inline citation tags, or use research mode which automatically includes citations in its answer.

Notes

• Timeout: Set client timeout to at least 30s for single-search, 300s (5 min) for research
• Single message: The messages array must contain exactly 1 user message
• Cost monitoring: Parse the <usage> tag from streaming responses to track costs