Skip to main content

MCP Integrations

Connect your Vectorize MCP agent to popular AI tools and development environments. Each integration takes just a few minutes to set up.

Available Integrations​

AI Assistants​

Claude Desktop​

Anthropic's desktop application with full MCP support.

  • Best for: General AI assistance with your data
  • Platform: macOS, Windows
  • Setup time: 2 minutes

Claude Code​

Terminal-based AI assistant from Anthropic.

  • Best for: Command-line workflows
  • Platform: macOS, Windows, Linux
  • Setup time: 1 minute

Code Editors​

Cursor​

AI-powered code editor built on VSCode.

  • Best for: AI-assisted coding with your documentation
  • Platform: macOS, Windows, Linux
  • Setup time: 2 minutes

Coming Soon​

  • Windsurf - Advanced AI code editor
  • Warp AI - Modern terminal with AI features
  • OpenAI Playground - Web-based testing environment
  • Groq Desktop - High-performance AI assistant

Integration Methods​

Vectorize supports two connection methods:

Most tools connect directly to Vectorize's MCP endpoint:

{
  "url": "https://agents.vectorize.io/api/agents/YOUR_AGENT_ID/mcp",
  "headers": {
    "Authorization": "Bearer YOUR_API_KEY"
  }
}

Advantages:

  • No local setup required
  • Automatic updates
  • Built-in monitoring
  • Higher reliability

2. Local MCP Server​

For development or special requirements:

{
  "command": "npx",
  "args": ["-y", "@vectorize-io/vectorize-mcp-server@latest"],
  "env": {
    "VECTORIZE_TOKEN": "YOUR_API_KEY"
  }
}

Use when:

  • Testing locally
  • Air-gapped environments
  • Custom modifications needed

Before You Begin​

Prerequisites​

✅ MCP Agent Created - You need an MCP agent in your workspace ✅ Functions Configured - At least one function must be set up ✅ API Key Generated - Create a key in your agent settings ✅ Tool Installed - Have your chosen AI tool ready

Quick Setup Checklist​

  1. Create MCP agent in Vectorize
  2. Configure at least one function
  3. Generate API key
  4. Copy provided configuration
  5. Add to your tool's settings
  6. Restart the tool
  7. Test with a query

Configuration Patterns​

Most MCP tools follow similar configuration patterns:

Desktop Applications​

  1. Locate configuration file (usually in app support folder)
  2. Add MCP server configuration
  3. Restart application

Command-Line Tools​

  1. Run setup command
  2. Provide API credentials
  3. Start using immediately

IDE Extensions​

  1. Open settings/preferences
  2. Navigate to MCP section
  3. Paste configuration
  4. Reload window

Security Best Practices​

API Key Management​

  • Create separate keys for each tool or user
  • Name keys descriptively (e.g., "John's Claude Desktop")
  • Rotate regularly - refresh keys monthly
  • Revoke unused keys immediately

Access Control​

  • Limit functions to necessary data only
  • Use hard-coded filters to restrict access
  • Monitor usage through agent dashboard
  • Review logs for unusual patterns

Network Security​

  • All connections use HTTPS/TLS
  • API keys are never logged
  • Rate limiting prevents abuse
  • IP allowlisting available (Enterprise)

Troubleshooting​

Common Issues​

Tool doesn't show MCP agent​

  1. Verify API key is correct
  2. Check tool was restarted after configuration
  3. Ensure agent has at least one function
  4. Confirm network connectivity

Authentication errors​

  1. Check API key hasn't expired
  2. Verify key belongs to correct agent
  3. Ensure proper header format
  4. Test key in Vectorize UI first

No results returned​

  1. Confirm pipeline has indexed documents
  2. Check function filters aren't too restrictive
  3. Test function in Vectorize UI
  4. Review function descriptions

Getting Help​

  • Documentation: Check tool-specific guides
  • Support: Contact support@vectorize.io
  • Community: Join our Discord server
  • Status: Check status.vectorize.io

Performance Tips​

Optimize Response Times​

  1. Use specific functions rather than general search
  2. Add metadata filters to reduce search space
  3. Index relevant metadata in your pipelines
  4. Cache frequently accessed data

Improve Result Quality​

  1. Write clear function descriptions
  2. Use descriptive parameter names
  3. Test functions before connecting tools
  4. Provide feedback on results

Usage Monitoring​

Track your MCP agent usage in the Vectorize dashboard:

  • Query volume - Requests per tool/function
  • Response times - Average latency
  • Error rates - Failed requests
  • Popular functions - Most-used tools

Next Steps​

Choose your integration:

Or continue learning:

Was this page helpful?