Exa API 配置指南

Your Configuration

Setting	Value
Coding Tool	Cursor
Framework	JavaScript
Use Case	Web search tool
Search Type	Auto - Balanced relevance and speed (~1 second)
Content	Compact
Project Description: (Not provided)

---

API Key Setup

Environment Variable

export EXA_API_KEY="YOUR_API_KEY"

.env File

EXA_API_KEY=YOUR_API_KEY

Usage in Code

import Exa from "exa-js";
 
const exa = new Exa(process.env.EXA_API_KEY);

---

🔌 Exa MCP Server for Cursor

Give Cursor real-time web search, code context, and company research with Exa MCP.

Add to Cursor settings (~/.cursor/mcp.json):

{
  "mcpServers": {
    "exa": {
      "type": "http",
      "url": "https://mcp.exa.ai/mcp?exaApiKey=YOUR_API_KEY",
      "headers": {}
    }
  }
}

Or use the one-click install button at docs.exa.ai/reference/exa-mcp

Available tools:

• web_search_exa - Real-time web search
• get_code_context_exa - Code snippets, docs, and examples
• company_research_exa - Company information and research
• crawling_exa - Extract content from any URL
• linkedin_search_exa - Find people on LinkedIn
• deep_researcher_start - AI-powered deep research

📖 Full docs: docs.exa.ai/reference/exa-mcp

---

Quick Start (JavaScript)

npm install exa-js

import Exa from "exa-js";
 
const exa = new Exa("YOUR_API_KEY");
 
const results = await exa.search("latest developments in AI safety research", {
  "type": "auto",
  "num_results": 10,
  "contents": {
    "highlights": {
      "max_characters": 4000
    }
  }
});
 
results.results.forEach(result => {
  console.log(result.title, result.url);
});

cURL (alternative)

curl -X POST 'https://api.exa.ai/search' \
  -H 'x-api-key: YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
  "query": "latest developments in AI safety research",
  "type": "auto",
  "num_results": 10,
  "contents": {
    "highlights": {
      "max_characters": 4000
    }
  }
}'

---

Function Calling / Tool Use

Function calling (also known as tool use) allows your AI agent to dynamically decide when to search the web based on the conversation context. Instead of searching on every request, the LLM intelligently determines when real-time information would improve its response—making your agent more efficient and accurate.

Why use function calling with Exa?

• Your agent can ground responses in current, factual information
• Reduces hallucinations by fetching real sources when needed
• Enables multi-step reasoning where the agent searches, analyzes, and responds

📚 Full documentation: https://docs.exa.ai/reference/openai-tool-calling

OpenAI Function Calling

import json
from openai import OpenAI
from exa_py import Exa
 
openai = OpenAI()
exa = Exa()
 
tools = [{
    "type": "function",
    "function": {
        "name": "exa_search",
        "description": "Search the web for current information.",
        "parameters": {
            "type": "object",
            "properties": {"query": {"type": "string", "description": "Search query"}},
            "required": ["query"]
        }
    }
}]
 
def exa_search(query: str) -> str:
    results = exa.search(query=query, type="auto", num_results=10, contents={"highlights": {"max_characters": 4000}})
    return "\n".join([f"{r.title}: {r.url}" for r in results.results])
 
messages = [{"role": "user", "content": "What's the latest in AI safety?"}]
response = openai.chat.completions.create(model="gpt-4o", messages=messages, tools=tools)
 
if response.choices[0].message.tool_calls:
    tool_call = response.choices[0].message.tool_calls[0]
    search_results = exa_search(json.loads(tool_call.function.arguments)["query"])
    messages.append(response.choices[0].message)
    messages.append({"role": "tool", "tool_call_id": tool_call.id, "content": search_results})
    final = openai.chat.completions.create(model="gpt-4o", messages=messages)
    print(final.choices[0].message.content)

Anthropic Tool Use

import anthropic
from exa_py import Exa
 
client = anthropic.Anthropic()
exa = Exa()
 
tools = [{
    "name": "exa_search",
    "description": "Search the web for current information.",
    "input_schema": {
        "type": "object",
        "properties": {"query": {"type": "string", "description": "Search query"}},
        "required": ["query"]
    }
}]
 
def exa_search(query: str) -> str:
    results = exa.search(query=query, type="auto", num_results=10, contents={"highlights": {"max_characters": 4000}})
    return "\n".join([f"{r.title}: {r.url}" for r in results.results])
 
messages = [{"role": "user", "content": "Latest quantum computing developments?"}]
response = client.messages.create(model="claude-sonnet-4-20250514", max_tokens=4096, tools=tools, messages=messages)
 
if response.stop_reason == "tool_use":
    tool_use = next(b for b in response.content if b.type == "tool_use")
    tool_result = exa_search(tool_use.input["query"])
    messages.append({"role": "assistant", "content": response.content})
    messages.append({"role": "user", "content": [{"type": "tool_result", "tool_use_id": tool_use.id, "content": tool_result}]})
    final = client.messages.create(model="claude-sonnet-4-20250514", max_tokens=4096, tools=tools, messages=messages)
    print(final.content[0].text)

---

Search Type Reference

Type	Best For	Speed	Depth
fast	Real-time apps, autocomplete, quick lookups	Fastest	Basic
auto	Most queries - balanced relevance & speed	Medium	Smart

Tip: type="auto" works well for most queries. It provides balanced relevance and speed.

---

Content Configuration

Choose ONE content type per request (not both):

Type	Config	Best For
Text	"text": {"max_characters": 20000}	Full content extraction, RAG
Highlights	"highlights": {"max_characters": 4000}	Snippets, summaries, lower cost

⚠️ Token usage warning: Using text: true (full page text) can significantly increase token count, leading to slower and more expensive LLM calls. To mitigate:

• Add max_characters limit: "text": {"max_characters": 10000}
• Use highlights instead if you don’t need contiguous text

When to use text vs highlights:

• Text - When you need untruncated, contiguous content (e.g., code snippets, full articles, documentation)
• Highlights - When you need key excerpts and don’t need the full context (e.g., summaries, Q&A, general research)

---

Domain Filtering (Optional)

Usually not needed - Exa’s neural search finds relevant results without domain restrictions.

When to use:

• Targeting specific authoritative sources
• Excluding low-quality domains from results

Example:

{
  "includeDomains": ["arxiv.org", "github.com"],
  "excludeDomains": ["pinterest.com"]
}

Note: includeDomains and excludeDomains cannot be used together.

---

Web Search Tool

Use category: "null" to search for null content.

{
  "query": "latest developments in AI safety research",
  "category": null,
  "num_results": 10,
  "contents": {
    "text": {
      "max_characters": 20000
    }
  }
}

Tips:

• Use type: "auto" for most queries
• Great for building search-powered chatbots or agents
• Combine with contents for RAG workflows

SDK Examples

---

Category Examples

Use category filters to search dedicated indexes. Each category returns only that content type.

Note: Categories can be restrictive. If you’re not getting enough results, try searching without a category first, then add one if needed.

People Search (category: "people")

Find people by role, expertise, or what they work on

await exa.search("software engineer distributed systems", {
  "category": "people",
  "type": "auto",
  "num_results": 10
});

Tips:

• Use SINGULAR form
• Describe what they work on
• No date/text filters supported

Company Search (category: "company")

Find companies by industry, criteria, or attributes

await exa.search("AI startup healthcare", {
  "category": "company",
  "type": "auto",
  "num_results": 10
});

Tips:

• Use SINGULAR form
• Simple entity queries
• Returns company objects, not articles

News Search (category: "news")

News articles

await exa.search("OpenAI announcements", {
  "category": "news",
  "type": "auto",
  "num_results": 10,
  "contents": {
    "text": {
      "max_characters": 20000
    }
  }
});

Tips:

• Use livecrawl: “preferred” for breaking news
• Avoid date filters unless required

Research Papers (category: "research paper")

Academic papers

await exa.search("transformer architecture improvements", {
  "category": "research paper",
  "type": "auto",
  "num_results": 10,
  "contents": {
    "text": {
      "max_characters": 20000
    }
  }
});

Tips:

• Use type: “auto” for most queries
• Includes arxiv.org, paperswithcode.com, and other academic sources

Tweet Search (category: "tweet")

Twitter/X posts

await exa.search("AI safety discussion", {
  "category": "tweet",
  "type": "auto",
  "num_results": 10,
  "contents": {
    "text": {
      "max_characters": 20000
    }
  }
});

Tips:

• Good for real-time discussions
• Captures public sentiment

---

Content Freshness (maxAgeHours)

maxAgeHours sets the maximum acceptable age (in hours) for cached content. If the cached version is older than this threshold, Exa will livecrawl the page to get fresh content.

Value	Behavior	Best For
24	Use cache if less than 24 hours old, otherwise livecrawl	Daily-fresh content
1	Use cache if less than 1 hour old, otherwise livecrawl	Near real-time data
0	Always livecrawl (ignore cache entirely)	Real-time data where cached content is unusable
-1	Never livecrawl (cache only)	Maximum speed, historical/static content
(omit)	Default behavior (livecrawl as fallback if no cache exists)	Recommended — balanced speed and freshness

When LiveCrawl Isn’t Necessary: Cached data is sufficient for many queries, especially for historical topics or educational content. These subjects rarely change, so reliable cached results can provide accurate information quickly.

See maxAgeHours docs for more details.

---

Other Endpoints

Beyond /search, Exa offers these endpoints:

Endpoint	Description	Docs
/contents	Get contents for known URLs	Docs
/answer	Q&A with citations from web search	Docs

Example - Get contents for URLs:

POST /contents
{
  "urls": ["https://example.com/article"],
  "text": { "max_characters": 20000 }
}

---

Troubleshooting

Results not relevant?

1. Try type: "auto" - most balanced option
2. Refine query - use singular form, be specific
3. Check category matches your use case

Results too slow?

1. Use type: "fast"
2. Reduce num_results
3. Skip contents if you only need URLs

No results?

1. Remove filters (date, domain restrictions)
2. Simplify query
3. Try type: "auto" - has fallback mechanisms

---

Resources

• Docs: https://exa.ai/docs
• Dashboard: https://dashboard.exa.ai
• API Status: https://status.exa.ai