Exa Search

Built-in Exa search tools for AI agents - neural search, content retrieval, similar pages, and AI answers

Prerequisites

• Python 3.10 or higher
• PraisonAI Agents package installed
• exa_py package installed
• EXA_API_KEY environment variable set

from praisonaiagents import Agent

agent = Agent(name="Researcher", tools=["exa_search"])
agent.start("Find recent papers on transformer architectures")

The user describes a research topic; Exa search returns high-quality web and paper results.

Exa provides AI-powered neural search capabilities optimized for LLM applications. PraisonAI includes built-in Exa tools for easy integration.

Quick Start

1. Install and set key

   pip install praisonaiagents exa_py
   export EXA_API_KEY=your_exa_api_key
   
2. Search with agent

   from praisonaiagents import Agent, exa
   
   agent = Agent(
       name="ResearchAgent",
       instructions="Search the web using Exa neural search.",
       tools=[exa],
   )
   
   agent.start("Find recent papers on transformer architectures")
   

Installation

pip install praisonaiagents exa_py

Setup

export EXA_API_KEY=your_exa_api_key
export OPENAI_API_KEY=your_openai_api_key

Built-in Exa Tool

PraisonAI provides a built-in exa tool that you can import directly:

from praisonaiagents import Agent
from praisonaiagents import exa

agent = Agent(
    name="SearchAgent",
    role="Web Researcher",
    goal="Find information on the web",
    tools=[exa]
)

result = agent.start("Find the hottest AI startups in 2025")
print(result)

Available Functions

Function	Description
exa	Neural search (alias for exa_search)
exa_search	Basic web search
exa_search_contents	Search with full text/highlights
exa_find_similar	Find similar pages to a URL
exa_answer	AI-generated answers with citations

Basic Usage

Simple Search

from praisonaiagents import exa

# Simple search
results = exa("AI startups 2025")
print(results)

Search with Options

from praisonaiagents import exa_search

results = exa_search(
    query="AI startups",
    num_results=10,
    type="neural",              # "auto", "neural", "fast", or "deep"
    category="company",         # Filter by category
    include_domains=["techcrunch.com", "wired.com"],
    start_published_date="2024-01-01"
)

for r in results.get("results", []):
    print(f"- {r['title']}: {r['url']}")

Search with Content

from praisonaiagents import exa_search_contents

results = exa_search_contents(
    query="AI in healthcare",
    text=True,           # Include full text
    highlights=True,     # Include relevant highlights
    num_results=5
)

for r in results.get("results", []):
    print(f"Title: {r['title']}")
    print(f"Text: {r.get('text', '')[:500]}...")
    if r.get('highlights'):
        print(f"Highlights: {r['highlights']}")

Find Similar Pages

from praisonaiagents import exa_find_similar

similar = exa_find_similar(
    url="https://openai.com",
    num_results=5,
    exclude_source_domain=True,  # Exclude openai.com from results
    category="company"
)

for r in similar.get("results", []):
    print(f"- {r['title']}: {r['url']}")

AI-Generated Answers

from praisonaiagents import exa_answer

result = exa_answer(
    query="What is the capital of France?",
    text=True  # Include citation text
)

print(f"Answer: {result['answer']}")
print(f"Citations: {len(result['citations'])}")
for c in result['citations']:
    print(f"  - {c['title']}: {c['url']}")

With PraisonAI Agent

from praisonaiagents import Agent
from praisonaiagents import exa

agent = Agent(
    name="ResearchAgent",
    role="AI Researcher",
    goal="Find and analyze information about AI companies",
    tools=[exa]
)

result = agent.start("Research the top 5 AI startups and their valuations")
print(result)

Using ExaTools Class

For more control, use the ExaTools class directly:

from praisonaiagents import ExaTools

# Initialize with custom API key (optional)
tools = ExaTools(api_key="your_api_key")  # or uses EXA_API_KEY env var

# Search
results = tools.search("AI news", num_results=5)

# Search with contents
results = tools.search_and_contents("AI healthcare", text=True, highlights=True)

# Find similar
similar = tools.find_similar("https://openai.com", num_results=5)

# Get answer
answer = tools.answer("What is GPT-4?", text=True)

Search Parameters

Parameter	Type	Description
query	str	Search query
num_results	int	Number of results (default 10, max 100)
type	str	”auto”, “neural”, “fast”, or “deep”
category	str	Data category filter
include_domains	list	Domains to include
exclude_domains	list	Domains to exclude
start_crawl_date	str	Only links crawled after this date
end_crawl_date	str	Only links crawled before this date
start_published_date	str	Only links published after this date
end_published_date	str	Only links published before this date
include_text	list	Strings that must be present
exclude_text	list	Strings that must not be present

Categories

Exa supports filtering by data categories:

Category	Description
company	Company websites
research paper	Academic papers
news	News articles
linkedin profile	LinkedIn profiles
github	GitHub repositories
tweet	Twitter/X posts
movie	Movie information
song	Music information
personal site	Personal websites
pdf	PDF documents
financial report	Financial reports

Deep Search

For comprehensive research, use deep search with additional queries:

from praisonaiagents import exa_search

results = exa_search(
    query="blog post about AI",
    type="deep",
    additional_queries=["AI blogpost", "machine learning blogs"],
    num_results=10
)

Structured Summaries

Get structured data from search results:

from praisonaiagents import ExaTools

tools = ExaTools()

company_schema = {
    "type": "object",
    "properties": {
        "name": {"type": "string"},
        "industry": {"type": "string"},
        "founded_year": {"type": "number"},
        "key_products": {"type": "array", "items": {"type": "string"}}
    },
    "required": ["name", "industry"]
}

results = tools.search_and_contents(
    query="OpenAI company information",
    summary={"schema": company_schema},
    category="company",
    num_results=3
)

# Parse structured summary
import json
for r in results.get("results", []):
    if r.get("summary"):
        data = json.loads(r["summary"])
        print(f"Company: {data.get('name')}")
        print(f"Industry: {data.get('industry')}")

Key Points

• Neural search: AI-powered semantic understanding of queries
• Environment variable: Set EXA_API_KEY before running
• Categories: Filter results by data type for cleaner results
• Deep search: Use additional queries for comprehensive research
• Structured output: Get JSON-formatted summaries with schemas

Best Practices

Use neural search for semantic queries

Exa excels at conceptual queries - use it over keyword search for research tasks.

Retrieve content in the same call

Set contents=True to get page content alongside search results, saving an extra step.

Filter by date for fresh results

Use start_published_date to filter for recent content when you need current information.

Use highlights for long documents

Enable highlights=True to get the most relevant passages instead of full page content.

Related

Custom Tools

Build your own agent tools

Tools Overview

Browse PraisonAI tool documentation