Automate in-depth web research and receive structured JSON results with citations.
The Research API is an asynchronous, multi-step pipeline that transforms open-ended questions into structured JSON answers backed by citations. You provide natural-language instructions (e.g. “Compare the hardware roadmaps of the top GPU manufacturers”) and an optional JSON Schema describing the output you want.
Under the hood, Exa agents perform multiple steps:
Planning – Your natural-language instructions
are parsed by an LLM that decomposes the task into one or more research steps.
Searching – Specialized search agents issue semantic and keyword queries to Exa’s search engine, continuously expanding and refining the result set until they can fulfil the request.
Reasoning & synthesis – Reasoning models combine facts across sources and return structured JSON (if you provide output.schema
) or a detailed markdown report.
Citations – Generates a citations
object mapping each root-level field to its supporting source passages so you can surface inline sources to your users.
Because tasks are asynchronous, you submit a request and immediately receive a task_id
. You can poll the task until it is complete or failed, or list all tasks to monitor progress in bulk.
The Research API offers two advanced agentic researcher models that break down your instructions, search the web, extract and reason over facts, and return structured answers with citations.
Typical completion times:
Model | p50 Time (seconds) | p90 Time (seconds) |
---|---|---|
exa-research | 20 | 40 |
exa-research-pro | 60 | 90 |
The Research API now uses variable usage-based pricing. You are billed based on how much work and reasoning the research agent does.
NOTE: You are ONLY charged for tasks that complete successfully.
Operation | exa-research | exa-research-pro | Notes |
---|---|---|---|
Search | $5/1,000 searches | $5/1,000 searches | Each unique search query issued by the agent |
Page read | $5/1,000 pages read | $10/1,000 pages read | One “page” = 1,000 tokens from the web |
Reasoning tokens | $5/1M tokens | $5/1M tokens | Specific LLM tokens used for reasoning and synthesis |
Example:
A research task with exa-research
that performs 6 searches, reads 20 pages of content, and uses 1,000 reasoning tokens would cost:
For exa-research-pro
, the same task would cost:
1. Competitive Landscape Table Goal: Compare the current flagship GPUs from NVIDIA, AMD, and Intel and extract pricing, TDP, and release date.
2. Market Size Estimate Goal: Estimate the total global market size (USD) for battery recycling in 2030 with a clear methodology.
3. Timeline of Key Events Goal: Build a timeline of major OpenAI product releases from 2015 – 2023.
Who is the Research API for?
Product teams, analysts, researchers, and anyone who needs structured answers that require reading multiple web sources — without having to build their own search + scraping + LLM pipeline.
How is this different from the /answer endpoint?
/answer
is designed for single-shot Q&A. The Research API handles
long-running, multi-step investigations. It’s suitable for tasks that
require complex reasoning over web data.
How long do tasks take?
Tasks generally complete in 20–40 seconds. Simple tasks that can be solved with few searches complete faster, while complex schema’s targeting niche subjects may take longer.
What are best practices for writing instructions?
Be explicit about the objective and any constraints
Specify the time range or types of sources to consult if important
Use imperative verbs (“Compare”, “List”, “Summarize”)
Keep it under 4096 characters
Are citations guaranteed?
Yes. Every root-level field in your data
object has a corresponding array of
citation URLs in the citations
object.
How large can my output schema be?
You must have ≤ 8 root fields. It must not be more than 5 fields deep.
What happens if my schema validation fails?
If your schema is not valid, an error will surface before the task is created with a message about what is invalid. You will not be charged for such requests.