Daraja API MCP Server

Daraja API MCP Server

Provides AI assistants with access to comprehensive Safaricom Daraja API documentation for all 22 M-Pesa APIs through searchable tools, enabling developers to query payment processing, transaction management, and business operations documentation.

Category
访问服务器

README

Daraja API Documentation Scraper & MCP Server

License: MIT Node.js Python Docker MCP

A comprehensive toolkit for accessing Safaricom's Daraja API documentation through web scraping and Model Context Protocol (MCP) integration. This project provides both a powerful scraper for keeping documentation up-to-date and a professional MCP server for AI assistant integration.

⚠️ Important Legal Notice & Disclaimers

Data Source and Usage Rights

  • Documentation Source: This project scrapes documentation from Safaricom's Daraja API portal
  • Authentication Required: The scraper requires login credentials to access Safaricom's developer portal
  • Terms of Service: Users must comply with Safaricom's Terms of Service when using the scraper
  • Data Ownership: All scraped documentation remains the intellectual property of Safaricom PLC

Legal Compliance

  • Personal Use: This tool is intended for personal development and learning purposes
  • Commercial Use: For commercial applications, ensure you have appropriate licenses from Safaricom
  • Rate Limiting: The scraper includes delays to respect Safaricom's servers - do not modify these
  • Account Responsibility: Users are responsible for their own Safaricom developer account credentials

Disclaimers

  • No Warranty: This software is provided "as is" without any warranties
  • Data Accuracy: Scraped documentation may become outdated - always verify with official sources
  • Service Availability: Safaricom may change their portal structure, potentially breaking the scraper
  • Legal Responsibility: Users assume all legal responsibility for their use of this tool

Ethical Usage Guidelines

  • Respect Rate Limits: Do not overwhelm Safaricom's servers with excessive requests
  • Valid Credentials: Only use your own legitimate Safaricom developer account
  • Data Sharing: Be mindful of sharing scraped data - respect Safaricom's intellectual property
  • Updates: Keep scraped documentation current and don't redistribute outdated information

By using this software, you acknowledge that you have read, understood, and agree to comply with these terms and Safaricom's Terms of Service.

Features

  • Complete Documentation Scraper: Automated scraping of all 22 Daraja APIs with images
  • Professional MCP Server: Standards-compliant server for AI assistant integration
  • Docker Support: Containerized deployment for consistent environments
  • Multi-Platform Support: Works on Windows, macOS, and Linux
  • Editor Integration: Compatible with Kiro, VSCode, Cursor, Windsurf, Claude Desktop, and more
  • Offline Ready: Complete documentation dataset included for immediate use
  • Auto-Discovery: Smart path detection for seamless setup

Table of Contents

Quick Start

Option 1: Docker Deployment (Recommended)

# Clone the repository
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp

# Build and start Docker container
cd mpesa-daraja-mcp
pnpm run docker:build
pnpm run docker:start

# Configure your editor (see Editor Integration section)

Option 2: Native Installation

# Clone the repository
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp

# Setup MCP server
cd mpesa-daraja-mcp
pnpm install
pnpm run build

# Configure your editor (see Editor Integration section)

Option 3: Fresh Scraping + MCP Setup

# Clone and setup scraper
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp

# Setup Python environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

# Run scraper (requires Daraja portal login)
python scraper.py

# Setup MCP server
cd mpesa-daraja-mcp
pnpm install
pnpm run build

Installation

Prerequisites

  • Docker (recommended) OR
  • Python 3.8+ (for scraper) + Node.js 18+ (for MCP server)
  • Git (for cloning)

Platform-Specific Setup

Windows

# Option 1: Docker (Recommended)
# Install Docker Desktop from https://docker.com
# Clone repository
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp\mpesa-daraja-mcp
pnpm run docker:build

# Option 2: Native Installation
# Install Python and Node.js from official websites
# Clone repository
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp

# Python setup (for scraper)
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

# Node.js setup (for MCP server)
cd mpesa-daraja-mcp
pnpm install
pnpm run build

macOS

# Option 1: Docker (Recommended)
# Install Docker Desktop from https://docker.com
brew install git
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp/mpesa-daraja-mcp
pnpm run docker:build

# Option 2: Native Installation
# Install dependencies via Homebrew
brew install python node git pnpm

# Clone repository
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp

# Python setup (for scraper)
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Node.js setup (for MCP server)
cd mpesa-daraja-mcp
pnpm install
pnpm run build

Linux (Ubuntu/Debian)

# Option 1: Docker (Recommended)
# Install Docker
sudo apt update
sudo apt install docker.io docker-compose git
sudo usermod -aG docker $USER
# Log out and back in, then:
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp/mpesa-daraja-mcp
pnpm run docker:build

# Option 2: Native Installation
# Install dependencies
sudo apt update
sudo apt install python3 python3-venv nodejs npm git
npm install -g pnpm

# Clone repository
git clone https://github.com/JacksCodeVault/mpesa-daraja-mcp.git
cd mpesa-daraja-mcp

# Python setup (for scraper)
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Node.js setup (for MCP server)
cd mpesa-daraja-mcp
pnpm install
pnpm run build

Docker Deployment

Docker provides the most reliable and consistent deployment method across all platforms.

Quick Docker Setup

cd mpesa-daraja-mcp

# Build the Docker image
pnpm run docker:build

# Start the container
pnpm run docker:start

# Check status
pnpm run docker:status

# View logs
pnpm run docker:logs

Docker Management Commands

# Build and deployment
pnpm run docker:build     # Build the Docker image
pnpm run docker:start     # Start the container
pnpm run docker:stop      # Stop the container
pnpm run docker:restart   # Restart the container

# Monitoring and debugging
pnpm run docker:status    # Check container status
pnpm run docker:logs      # View container logs
pnpm run docker:shell     # Open shell in container

# Cleanup
pnpm run docker:clean     # Remove container and image

Docker Configuration

The Docker setup includes:

  • Multi-stage build for optimized image size
  • Non-root user for security
  • Volume mounting for documentation access
  • Health checks for reliability
  • Resource limits for production use

Docker Compose Configuration

# docker-compose.yml
services:
  daraja-mcp:
    build: .
    container_name: daraja-mcp-server
    restart: unless-stopped
    volumes:
      - ../daraja_docs_v3:/app/daraja_docs_v3:ro
    environment:
      - NODE_ENV=production
    stdin_open: true
    tty: true

MCP Configuration for Docker

When using Docker, update your editor's MCP configuration:

{
  "mcpServers": {
    "daraja-docs": {
      "command": "docker",
      "args": ["exec", "-i", "daraja-mcp-server", "node", "dist/index.js"],
      "disabled": false,
      "autoApprove": [
        "search_daraja_apis",
        "get_daraja_api_doc",
        "list_apis_by_category",
        "get_api_summary",
        "get_server_stats",
        "compare_apis"
      ]
    }
  }
}

Scraper Usage

⚠️ IMPORTANT: Before using the scraper, ensure you have:

  • A valid Safaricom developer account
  • Accepted Safaricom's Terms of Service
  • Permission to access the documentation programmatically
  • Understanding that you're responsible for compliance with Safaricom's policies

The scraper automatically downloads documentation for all 22 Daraja APIs with images.

First Time Setup

  1. Activate Python environment:

    source venv/bin/activate  # Linux/macOS
    venv\Scripts\activate     # Windows
    
  2. Run the scraper:

    python scraper.py
    
  3. Login Process:

    • Browser window opens automatically
    • Login to Daraja portal manually when prompted
    • Press ENTER in terminal after successful login
    • Scraper saves session for future use

Updating Documentation

To get the latest API documentation:

# Activate environment
source venv/bin/activate

# Run scraper (uses saved session)
python scraper.py

# If session expired, login again when browser opens

Output Structure

daraja_docs_v3/
├── data_index.json          # API metadata and index
├── docs/                    # Markdown documentation
│   ├── Authorization.md
│   ├── MpesaExpressSimulate.md
│   └── ... (22 API files)
└── images/                  # Downloaded images
    ├── Authorization_img_0.svg
    └── ... (200+ images)

MCP Server Setup

The MCP server provides AI assistants with access to Daraja documentation through 6 powerful tools.

Native Build and Test

cd mpesa-daraja-mcp

# Install dependencies
pnpm install

# Build TypeScript
pnpm run build

# Test server can find documentation
pnpm test

# Run server (for testing)
pnpm run dev

Available MCP Tools

The MCP server provides 6 comprehensive tools:

  1. search_daraja_apis - Advanced search with category filtering
  2. get_daraja_api_doc - Complete API documentation retrieval
  3. list_apis_by_category - Organized API listing by categories
  4. get_api_summary - Enhanced summaries with endpoints
  5. get_server_stats - Usage statistics and monitoring
  6. compare_apis - Side-by-side API comparison

Server Configuration

The server automatically detects documentation in these locations:

  • ../daraja_docs_v3 (relative to server)
  • ./daraja_docs_v3 (current directory)
  • /app/daraja_docs_v3 (Docker container)
  • Custom path via environment variable

Editor Integration

Kiro IDE

  1. Create MCP configuration file:

    # Create .kiro/settings/mcp.json in your workspace
    mkdir -p .kiro/settings
    
  2. Docker Configuration (Recommended):

    {
      "mcpServers": {
        "daraja-docs": {
          "command": "docker",
          "args": ["exec", "-i", "daraja-mcp-server", "node", "dist/index.js"],
          "disabled": false,
          "autoApprove": [
            "search_daraja_apis",
            "get_daraja_api_doc",
            "list_apis_by_category",
            "get_api_summary",
            "get_server_stats",
            "compare_apis"
          ]
        }
      }
    }
    
  3. Native Configuration:

    {
      "mcpServers": {
        "daraja-docs": {
          "command": "node",
          "args": ["mpesa-daraja-mcp/dist/index.js"],
          "disabled": false,
          "autoApprove": [
            "search_daraja_apis",
            "get_daraja_api_doc",
            "list_apis_by_category",
            "get_api_summary",
            "get_server_stats",
            "compare_apis"
          ]
        }
      }
    }
    

VSCode with MCP Extensions

  1. Install MCP extension (if available)

  2. Configure in settings.json:

    {
      "mcp.servers": {
        "daraja-docs": {
          "command": "docker",
          "args": ["exec", "-i", "daraja-mcp-server", "node", "dist/index.js"]
        }
      }
    }
    

Cursor IDE

  1. Open Cursor settings

  2. Add MCP server configuration:

    {
      "mcp": {
        "servers": {
          "daraja-docs": {
            "command": "docker",
            "args": ["exec", "-i", "daraja-mcp-server", "node", "dist/index.js"]
          }
        }
      }
    }
    

Windsurf

  1. Access Windsurf MCP settings

  2. Add server configuration:

    {
      "mcpServers": {
        "daraja-docs": {
          "command": "docker",
          "args": ["exec", "-i", "daraja-mcp-server", "node", "dist/index.js"]
        }
      }
    }
    

Claude Desktop

  1. Edit Claude config file:

    • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
    • Windows: %APPDATA%\Claude\claude_desktop_config.json
  2. Add server configuration:

    {
      "mcpServers": {
        "daraja-docs": {
          "command": "docker",
          "args": ["exec", "-i", "daraja-mcp-server", "node", "dist/index.js"]
        }
      }
    }
    

Generic MCP Client

For any MCP-compatible client using Docker:

{
  "servers": {
    "daraja-docs": {
      "command": "docker",
      "args": ["exec", "-i", "daraja-mcp-server", "node", "dist/index.js"],
      "env": {}
    }
  }
}

For native installation:

{
  "servers": {
    "daraja-docs": {
      "command": "node",
      "args": ["dist/index.js"],
      "cwd": "/absolute/path/to/mpesa-daraja-mcp",
      "env": {}
    }
  }
}

API Documentation

Available APIs

The project includes comprehensive documentation for 22 Daraja APIs organized into 6 categories:

Core Payment APIs

  • Authorization - OAuth token generation and management
  • MpesaExpressSimulate - STK Push payment initiation
  • MpesaExpressQuery - STK Push transaction status queries

Payment Processing

  • CustomerToBusiness - C2B payment processing
  • BusinessToCustomer - B2C payment disbursements
  • CustomerToBusinessRegisterURL - C2B URL registration

Transaction Management

  • TransactionStatus - Transaction status verification
  • AccountBalance - Account balance inquiries
  • Reversal - Transaction reversal operations
  • PullTransaction - Transaction detail retrieval

Business Operations

  • BusinessPayBill - Bill payment processing
  • BusinessBuyGoods - Goods purchase payments
  • B2BExpressCheckout - Business-to-business transactions
  • BusinessToPochi - Business to Pochi wallet transfers

Advanced Features

  • DynamicQRCode - Dynamic QR code generation
  • BillManager - Bill management and processing
  • TaxRemittance - Tax payment processing
  • MpesaRatiba - Scheduled payment management

Specialized Services

  • B2CAccountTopUp - Account top-up services
  • Swap - Currency swap operations
  • IMSI - SIM card management services
  • IotSimManagement - IoT SIM card management

Using the MCP Tools

Advanced API Search

Use search_daraja_apis tool with powerful filtering:

  • Query: "payment" + Category: "core" → Core payment APIs
  • Query: "balance" → AccountBalance and related APIs
  • Category: "business" → All business operation APIs
  • No parameters → All 22 APIs with categories

Complete Documentation Access

Use get_daraja_api_doc tool for full documentation:

  • api_name: "Authorization" → Complete OAuth implementation guide
  • api_name: "MpesaExpressSimulate" → Full STK Push documentation
  • Includes code examples, parameters, and response formats

Enhanced API Summaries

Use get_api_summary tool for quick overviews:

  • api_name: "BusinessToCustomer" → Summary with key endpoints
  • include_endpoints: true → Includes endpoint URLs
  • Perfect for quick reference and comparison

API Comparison

Use compare_apis tool for side-by-side analysis:

  • api_names: ["Authorization", "MpesaExpressSimulate"] → Compare auth vs payment
  • comparison_aspects: ["endpoints", "authentication"] → Focus on specific aspects
  • Supports 2-4 APIs simultaneously

Usage Statistics

Use get_server_stats tool for monitoring:

  • Total API count and categories
  • Most accessed APIs
  • Request statistics
  • Server health information

Category-Based Browsing

Use list_apis_by_category tool for organized access:

  • category: "payments" → All payment-related APIs
  • No category → List all categories with counts
  • Perfect for discovering related APIs

Troubleshooting

Docker Issues

Container Won't Start

# Check Docker is running
docker --version
docker-compose --version

# View detailed logs
pnpm run docker:logs

# Rebuild if needed
pnpm run docker:clean
pnpm run docker:build

Documentation Not Found in Container

# Check volume mounting
docker inspect daraja-mcp-server

# Verify documentation exists
ls -la ../daraja_docs_v3/data_index.json

# Test container access
pnpm run docker:shell
ls -la /app/daraja_docs_v3/

MCP Connection Issues with Docker

# Ensure container is running
pnpm run docker:status

# Test MCP tools directly
docker exec -i daraja-mcp-server node dist/index.js

# Check container logs for errors
pnpm run docker:logs

Native Installation Issues

MCP Server Issues

Server won't start:

# Check Node.js version
node --version  # Should be 18+

# Rebuild server
cd mpesa-daraja-mcp
pnpm run clean
pnpm install
pnpm run build

Documentation not found:

# Test documentation detection
pnpm test

# Check paths manually
ls -la ../daraja_docs_v3/data_index.json

TypeScript errors:

# Update dependencies
pnpm update
pnpm run build

# Check for syntax errors
pnpm run dev

Scraper Issues

Browser doesn't open:

# Install browser dependencies
playwright install chromium

# Check Python version
python --version  # Should be 3.8+

Login session expired:

# Delete auth file and re-login
rm auth.json
python scraper.py

Permission errors:

# Check file permissions
chmod +x scraper.py
# Or run with python explicitly
python scraper.py

Editor Integration Issues

MCP Server Not Connecting

  1. Verify Docker container is running:

    pnpm run docker:status
    
  2. Check MCP configuration syntax:

    • Ensure JSON is valid
    • Verify command and args are correct
    • Use absolute paths for native installation
  3. Test server manually:

    # Docker
    docker exec -i daraja-mcp-server node dist/index.js
    
    # Native
    cd mpesa-daraja-mcp
    node dist/index.js
    
  4. Restart editor after configuration changes

Tools Not Appearing

  1. Check server logs in editor's MCP panel
  2. Verify autoApprove list includes desired tools
  3. Ensure server builds successfully: pnpm run build
  4. Test with minimal configuration first

Platform-Specific Issues

Windows Platform Issues

  • Use forward slashes in Docker paths: C:/path/to/project
  • Run PowerShell as Administrator if Docker permission issues
  • Check Windows Defender isn't blocking Docker or Node.js
  • Use WSL2 for better Docker performance

macOS Platform Issues

  • Install Xcode Command Line Tools: xcode-select --install
  • Use Homebrew for dependency management: brew install docker
  • Check Docker Desktop is running and configured
  • Verify file permissions: chmod +x docker-scripts.sh

Linux Platform Issues

  • Add user to docker group: sudo usermod -aG docker $USER
  • Install build essentials: sudo apt install build-essential
  • Check Docker service: sudo systemctl status docker
  • Verify Node.js installation: which node

Performance Optimization

Docker Performance

# Limit container resources
docker update --memory=512m --cpus="0.5" daraja-mcp-server

# Use multi-stage builds for smaller images
# (Already implemented in Dockerfile)

# Clean up unused Docker resources
docker system prune -a

MCP Server Performance

# Monitor server stats
# Use get_server_stats tool to track usage

# Optimize documentation loading
# Server caches documentation on startup

# Use appropriate log levels
NODE_ENV=production pnpm run docker:start

Getting Help

  1. Check logs first: Both Docker and native installations provide detailed error messages
  2. Test components individually: Use pnpm test to verify MCP server setup
  3. Verify paths and permissions: Ensure all file paths are correct and accessible
  4. Update dependencies: Keep Docker, Node.js, and Python packages current
  5. Use Docker for consistency: Docker eliminates most platform-specific issues

Contributing

⚠️ Data Handling Guidelines for Contributors:

  • No Scraped Data in PRs: Do not include scraped documentation in pull requests
  • Respect IP Rights: Ensure contributions don't violate Safaricom's intellectual property
  • Code Only: Contribute improvements to the scraper and MCP server code, not the data
  • Documentation: Update README and code comments, not scraped API documentation

Development Setup

  1. Fork the repository

  2. Create feature branch: git checkout -b feature/amazing-feature

  3. Setup development environment:

    # Docker development (recommended)
    cd mpesa-daraja-mcp
    pnpm run docker:build
    pnpm run docker:start
    
    # Native development
    # Python development
    pip install -r requirements-dev.txt  # If exists
    
    # Node.js development
    cd mpesa-daraja-mcp
    pnpm install
    pnpm run dev
    

Adding New APIs

  1. Update scraper URLs:

    URLS = [
        "https://developer.safaricom.co.ke/apis/NewAPI",
        # ... existing URLs
    ]
    
  2. Test scraping:

    python scraper.py
    
  3. Update API categorization:

    // In mpesa-daraja-mcp/src/config.ts
    API_CATEGORIES: {
      new_category: ["NewAPI"],
      // ... existing categories
    }
    
  4. Verify MCP server:

    cd mpesa-daraja-mcp
    pnpm test
    pnpm run docker:build
    

Code Style

  • Python: Follow PEP 8, use black formatter
  • TypeScript: Use Prettier formatting, ESLint rules
  • Docker: Follow best practices for multi-stage builds
  • Commits: Use conventional commit format

Testing

# Test scraper
python scraper.py --test-mode

# Test MCP server (native)
cd mpesa-daraja-mcp
pnpm test
pnpm run build

# Test MCP server (Docker)
pnpm run docker:build
pnpm run docker:start
pnpm run docker:logs

Documentation

  • Update README for new features
  • Add inline code comments
  • Update API documentation
  • Include Docker-specific instructions

License

This project is licensed under the MIT License - see the LICENSE file for details.

⚠️ Important License Clarification:

  • Software License: The MIT License applies only to the scraper and MCP server code
  • Documentation Rights: Scraped Daraja API documentation remains the property of Safaricom PLC
  • Separate Terms: Use of Safaricom's documentation is subject to their Terms of Service
  • No Transfer: This license does not grant rights to Safaricom's intellectual property

What this means

  • Commercial use - Use this project in commercial applications
  • Modification - Modify the source code to fit your needs
  • Distribution - Distribute copies of the software
  • Private use - Use the software for private purposes
  • Liability - The authors are not liable for any damages
  • Warranty - The software is provided "as is" without warranty

Attribution

If you use this project, please consider:

  • Starring the repository on GitHub
  • 📝 Mentioning the project in your documentation
  • 🔗 Linking back to the original repository

Acknowledgments

  • Safaricom for providing the Daraja API platform
  • Model Context Protocol team for the MCP standard
  • Playwright team for the excellent automation framework
  • Docker community for containerization best practices
  • Open source contributors who make projects like this possible

Repository Information

Support

If you find this project helpful, please consider:

  • Starring the repository
  • 🐛 Reporting issues you encounter
  • 💡 Suggesting improvements
  • 🤝 Contributing to the project
  • 📢 Sharing with others who might benefit

Ready to integrate Daraja API documentation with your AI assistant?

Quick Start with Docker: cd mpesa-daraja-mcp && pnpm run docker:build && pnpm run docker:start

Browse the documentation: Use the MCP tools to explore 22 comprehensive Daraja APIs

Need help? Check the Troubleshooting section or open an issue

Start with the Quick Start guide and have your AI assistant accessing Daraja documentation in minutes!

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选