Skip to main content
POST
/
v1
/
deep-research
curl --request POST \
  --url https://api.incredible.one/v1/deep-research \
  --header 'Content-Type: application/json' \
  --data '{
  "instructions": "Research the latest developments in quantum computing"
}'
{
  "success": true,
  "research_id": "research-abc123",
  "status": "completed",
  "output": {
    "gpus": [
      {
        "manufacturer": "NVIDIA",
        "model": "RTX 4090",
        "price": 1599,
        "tdp": 450,
        "launchDate": "2022-10-12"
      }
    ]
  },
  "citations": [
    {
      "url": "https://nvidia.com/specs",
      "title": "NVIDIA RTX 4090 Specifications"
    }
  ],
  "searches_performed": 8,
  "pages_read": 42
}

​
Overview

Perform comprehensive, multi-step research that:
  • Plans research — Determines what searches to run
  • Gathers information — Performs multiple web searches
  • Synthesizes findings — Combines information into a report
  • Provides citations — Links back to original sources
  • Structured output — Returns JSON matching your schema
This is like having an AI research assistant that can gather, analyze, and format information from across the web.

​
How it works

  1. You provide: Natural language instructions + optional JSON schema
  2. Exa researches: Plans searches, reads pages, extracts information
  3. You receive: Structured report with citations and metadata

​
Models

​
exa-research (default)

Standard research model for most tasks:
  • Fast results
  • Good quality
  • Cost-effective

​
exa-research-pro

Enhanced reasoning for complex tasks:
  • Better analysis
  • More thorough research
  • Higher quality synthesis

​
Output schemas

Define exactly how you want results formatted:
Company comparison
{
  "type": "object",
  "properties": {
    "companies": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "name": {"type": "string"},
          "founded": {"type": "string"},
          "revenue": {"type": "number"},
          "employees": {"type": "integer"}
        }
      }
    }
  }
}
Market analysis
{
  "type": "object",
  "properties": {
    "market_size": {"type": "number"},
    "growth_rate": {"type": "number"},
    "key_players": {
      "type": "array",
      "items": {"type": "string"}
    },
    "trends": {
      "type": "array",
      "items": {"type": "string"}
    }
  }
}
Technical comparison
{
  "type": "object",
  "properties": {
    "features": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "name": {"type": "string"},
          "supported_by": {
            "type": "array",
            "items": {"type": "string"}
          }
        }
      }
    }
  }
}

​
Metadata

Each response includes useful metadata:
  • research_id — Unique identifier for this research task
  • status — Completion status
  • searches_performed — Number of searches executed
  • pages_read — Number of web pages analyzed
  • citations — Source URLs with titles

​
Use cases

Competitive Intelligence
  • Compare competitors’ features and pricing
  • Analyze market positioning
  • Track product launches
Market Research
  • Industry size and growth
  • Key players and trends
  • Customer sentiment
Technical Research
  • Technology comparisons
  • Best practices and patterns
  • Implementation guides
Due Diligence
  • Company backgrounds
  • Financial information
  • Leadership profiles
Content Creation
  • Research for articles
  • Data for presentations
  • Fact-checking sources

​
Caching

Enable caching to reuse results for identical requests: 
{
  "instructions": "...",
  "use_cache": true
}
Benefits:
  • Faster responses
  • Lower costs
  • Consistent results
Cache duration: 24 hours

​
Pricing notes

Deep Research uses Exa.AI. Costs vary based on:
  • Number of searches performed
  • Number of pages read
  • Reasoning tokens used
  • Model selected (pro costs more)
Set EXA_API_KEY environment variable to use this endpoint.

​
Tips for best results

  1. Be specific: Clear instructions get better results
  2. Use schemas: Structured output ensures consistency
  3. Include context: Mention time frames, regions, or specific aspects
  4. Request citations: Ask for sources in your instructions
  5. Choose the right model: Use pro for complex analysis

​
Example instructions

Good ✅
“Compare the flagship smartphones from Apple, Samsung, and Google released in 2024. Include model name, starting price, screen size, camera specs, and battery capacity. Provide citations for each specification.”
Too vague ❌
“Tell me about phones”
Good ✅
“Research the current state of quantum computing. Focus on: 1) Leading companies and their qubit counts, 2) Recent breakthroughs in 2024, 3) Expected timeline for commercial applications. Return as structured data with citations.”
Too vague ❌
“What’s happening with quantum computing?”

​
Next steps

  • Web search: /api-reference/services/web-search
  • Answer endpoint: /api-reference/services/answer
  • Chat completion: /api-reference/chat/completion

Body

application/json
​
instructions
string
required

Research task instructions in natural language

​
output_schema
object

JSON schema for structured output

​
model
enum<string>
default:exa-research

Model to use: 'exa-research' (standard) or 'exa-research-pro' (enhanced reasoning)

Available options:
exa-research,
exa-research-pro
​
use_cache
boolean
default:true

Use cached results for identical requests

Response

Research completed successfully

​
success
boolean
required

Whether the request was successful

​
research_id
string
required

Unique identifier for this research task

​
status
string
required

Completion status

​
output
object
required

Structured research output

​
citations
object[]

Source citations

​
searches_performed
integer

Number of searches executed

​
pages_read
integer

Number of web pages analyzed