Exa Search Skill

Semantic search skill for Claude Code using Exa API. Provides embeddings-based intelligent search, similar content discovery, and structured research capabilities.

Features

• 🔍 Semantic Search: Neural, fast, deep, and auto search modes
• 📄 Content Extraction: Get clean, parsed HTML content from search results
• 🔗 Find Similar: Discover semantically similar pages based on a URL
• 💬 Direct Answers: Get immediate answers to questions
• 📊 Structured Research: Automated research with custom output schemas
• 🎯 Category Filtering: Search by company, people, research papers, news, PDFs, GitHub, tweets, etc.
• 💰 Cost Tracking: Detailed cost breakdown for each request

Installation

Option 1: Install via skills CLI (Recommended)

# Install globally to all detected agents (Claude Code, Cursor, Codex, etc.)
npx skills add -g BenedictKing/exa-search

# Or install to current project only
npx skills add BenedictKing/exa-search

The skill will be automatically installed and loaded by Claude Code.

Option 2: Manual Installation via Git Clone

1. Clone this repository to your Claude skills directory:

cd ~/.claude/skills
git clone https://github.com/BenedictKing/exa-search.git

2. Configure your API key:

cd exa-search/.claude/skills/exa-search
cp .env.example .env
# Edit .env and add your Exa API key

Configuration

Get your API key from Exa Dashboard.

Two ways to configure:

1. Environment Variable (recommended):

export EXA_API_KEY=your_api_key_here

2. .env file:

# In .claude/skills/exa-search/.env
EXA_API_KEY=your_api_key_here

Usage

Trigger the Skill

Use /exa-search or let Claude automatically invoke it when you need:

• Semantic web search
• Finding similar content
• Direct answers to questions
• Structured research

Example Queries

Semantic Search:

Find latest research papers on transformer architectures

Find Similar:

Find pages similar to https://arxiv.org/abs/1706.03762

Direct Answer:

What are the key differences between GPT-4 and Claude?

Structured Research:

Research the top AI companies and their key metrics

API Endpoints

1. Search

Embeddings-based semantic search with multiple modes:

• neural: Semantic search using embeddings
• fast: Quick keyword-based search
• auto: Automatically choose best method (default)
• deep: Comprehensive deep search

2. Contents

Extract full content from search result IDs.

3. Find Similar

Discover semantically similar pages based on a given URL.

4. Answer

Get direct answers to questions with source citations.

5. Research

Automated deep research with custom output schemas.

Advanced Features

Category Filtering

Search within specific content types:

• company, people, research paper, news, pdf, github, tweet, etc.

Date Filtering

Filter by publication date or crawl date:

{
  "startPublishedDate": "2025-01-01",
  "endPublishedDate": "2025-12-31"
}

Domain Control

Include or exclude specific domains:

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

Content Options

Control what content to retrieve:

{
  "contents": {
    "text": true,
    "highlights": true,
    "summary": true
  }
}

Pricing

Exa API pricing (as of 2026):

• Neural Search (1-25 results): $0.005
• Neural Search (26-100 results): $0.025
• Deep Search (1-25 results): $0.015
• Deep Search (26-100 results): $0.075
• Content text/highlights/summary: $0.001 per page

Each API response includes a costDollars field with detailed cost breakdown.

Architecture

This skill uses a two-phase architecture:

1. Main skill: Understands user intent, chooses endpoint, assembles payload
2. Sub-skill (exa-fetcher): Executes HTTP calls in isolated context

This design minimizes token usage and keeps conversation history clean.

Comparison with Tavily

| Feature | Exa | Tavily | |---------|-----|--------| | Search Type | Embeddings-based semantic | Keyword + AI-enhanced | | Find Similar | ✅ | ❌ | | Category Search | ✅ (company, people, papers, etc.) | ❌ | | Search Modes | neural/fast/deep/auto | basic/advanced | | Cost Tracking | ✅ Detailed per request | ❌ | | Crawl/Map | ❌ | ✅ |

Use Exa when:

• You need semantic/embeddings-based search
• Finding similar content is important
• You want category-specific searches
• Cost tracking is required

Use Tavily when:

• You need site crawling/mapping
• You want structured research with citations
• You prefer keyword-based search

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT License - see LICENSE file for details.

Author

BenedictKing

• GitHub: @BenedictKing

Links

• Exa API Documentation
• Exa Dashboard
• Claude Code