Skip to main content
New to Exa? Try the Coding Agent Quickstart to get started in under a minute.
The official Python SDK for Exa. Search the web, get page contents, find similar pages, and get answers with citations.

Get API Key

Get your API key from the dashboard

​
Install

bash pip install exa-py
Requires Python 3.9+

​
Quick Start

from exa_py import Exa

exa = Exa()  # reads EXA_API_KEY from environment
Search the web and get page contents in one call. 
results = exa.search(
    "blog post about artificial intelligence",
    contents={"highlights": {"max_characters": 4000}}
)

results = exa.search(
    "climate tech news",
    num_results=20,
    start_published_date="2024-01-01",
    include_domains=["techcrunch.com", "wired.com"],
    contents={"highlights": {"max_characters": 4000}}
)

structured_results = exa.search(
    "Who is the CEO of OpenAI?",
    type="deep",
    system_prompt="Prefer official sources and avoid duplicate results",
    output_schema={
        "type": "object",
        "properties": {
            "leader": {"type": "string"},
            "title": {"type": "string"},
            "source_count": {"type": "number"}
        },
        "required": ["leader", "title"]
    },
    contents={"highlights": {"max_characters": 4000}}
)

print(structured_results.output.content if structured_results.output else None)
output_schema and system_prompt work across all search types. Keep output_schema focused on the fields you want in output.content, and use system_prompt to guide the final returned result. For more demanding synthesis, prefer reasoning-focused search types like deep-lite, deep, deep-reasoning, or deep-max. Do not add citations/confidence fields there; grounding is returned automatically in output.grounding. Adding citation/confidence fields to output_schema creates duplicate data, weaker structure, and less reliable attribution.
Reasoning-focused search variants:
  • deep-lite: lowest-latency deep-search mode
  • deep: light mode
  • deep-reasoning: base reasoning mode
  • deep-max: maximum-effort deep-search mode
Need streaming structured synthesis? The raw /search endpoint supports stream: true together with outputSchema and returns OpenAI-compatible SSE chunks. See the Search API guide.

​
Get Contents

Get text, summaries, or highlights from URLs. 
results = exa.get_contents(
    ["https://openai.com/research"],
    text=True
)

results = exa.get_contents(
    ["https://stripe.com/docs/api"],
    summary=True
)

results = exa.get_contents(
    ["https://arxiv.org/abs/2303.08774"],
    highlights={"max_characters": 2000}
)

​
Find Similar

Find pages similar to a URL. 
results = exa.find_similar(
    "https://paulgraham.com/greatwork.html",
    contents={"text": True}
)

results = exa.find_similar(
    "https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html",
    exclude_source_domain=True,
    contents={"text": True}
)

​
Answer

Get answers to questions with citations. 
response = exa.answer("What caused the 2008 financial crisis?")
print(response.answer)

for chunk in exa.stream_answer("Explain quantum computing"):
    print(chunk, end="", flush=True)

​
Async

Use AsyncExa for async operations. 
from exa_py import AsyncExa

exa = AsyncExa()

results = await exa.search(
    "machine learning startups",
    contents={"highlights": {"max_characters": 4000}}
)

​
Research

Run research tasks with structured output. 
task = exa.research.create(
    instructions="Summarize recent advances in fusion energy",
    output_schema={
        "type": "object",
        "properties": {
            "summary": {"type": "string"},
            "key_developments": {"type": "array", "items": {"type": "string"}}
        }
    }
)

result = exa.research.poll_until_finished(task.research_id)

​
Resources