Installing MCP Servers
Learn how to install and configure third-party MCP servers to extend your AI assistant's capabilities beyond the built-in Voka AI integrations.
What Are Third-Party MCP Servers?
Third-party MCP servers are community-built or vendor-provided integrations that follow the MCP protocol. They allow your assistant to connect to platforms not natively supported by Voka AI's MCP server.
Examples:
- Database connections (PostgreSQL, MySQL)
- Custom business tools
- Specialized APIs
- Internal company systems
Before You Install
You'll need:
- Access to the Voka AI dashboard
- MCP server URL or package name
- Any required API keys or credentials
- Basic understanding of the platform you're integrating
Note: Most users don't need third-party MCP servers. The built-in Voka AI integrations (Acuity, Square, Jobber, etc.) work out-of-the-box with no installation required.
Installation Methods
Method 1: Pre-Built MCP Server URL
If you have an MCP server URL (e.g., from a vendor or community repository):
- Go to AI Assistants in your dashboard
- Edit the assistant you want to configure
- Click the Agent tab
- Scroll to MCP Servers section
- Click Add MCP Server
- Enter:
- Name: Descriptive name (e.g., "Company Database")
- URL: Full MCP server URL (e.g.,
https://mcp.example.com) - API Key: If required by the server
- Click Save
- Test the connection
Method 2: NPM Package Installation
Some MCP servers are distributed as NPM packages:
- Find the package name (e.g.,
@company/mcp-server) - In Voka AI dashboard:
- Go to AI Assistants → Edit assistant
- Agent tab → MCP Servers
- Click Add MCP Server
- Enter package name:
@company/mcp-server - Configure any required environment variables
- Click Save
Voka AI will install and run the package automatically.
Configuration
After installation, configure how your assistant uses the MCP server:
Enable/Disable
Toggle the MCP server on/off for specific assistants:
- Enabled: Assistant can query this server
- Disabled: Server installed but not active
Environment Variables
Some MCP servers require configuration:
API_KEY=your_api_key_here
DATABASE_URL=postgresql://user:pass@host:5432/db
REGION=us-west-2
Add these in the MCP Server settings panel.
Permissions
Control what actions the MCP server can perform:
- Read-only: Query data only
- Read-write: Query and modify data
- Full access: All operations
Testing Your Installation
After installing an MCP server:
- Make a test call to your assistant
- Ask a question that requires the MCP server
- Check the call logs for:
- MCP server queries
- Successful responses
- Any error messages
- Review transcript to verify correct behavior
Example test:
You: "What's in our database for customer ID 12345?"
Assistant: [Queries MCP server]
Assistant: "Customer 12345 is John Smith, account created March 2024..."
Common MCP Servers
Database Servers
- PostgreSQL MCP - Query PostgreSQL databases
- MySQL MCP - Connect to MySQL databases
- MongoDB MCP - Access MongoDB collections
Business Tools
- Salesforce MCP - CRM data access
- QuickBooks MCP - Financial data queries
- Shopify MCP - E-commerce integration
Utilities
- Email MCP - Send emails via SMTP
- SMS MCP - Send text messages
- File Storage MCP - Access cloud storage (S3, Google Drive)
Find more: Browse the MCP Server Registry for community-built servers.
Troubleshooting
"Connection Failed"
- Verify the MCP server URL is correct
- Check that API keys are valid
- Ensure the server is online and accessible
"Authentication Error"
- Double-check API credentials
- Verify environment variables are set correctly
- Check for expired tokens
"Server Not Responding"
- Test the MCP server URL directly (if possible)
- Check server status/uptime
- Review server logs for errors
"Permission Denied"
- Ensure your API key has required permissions
- Check MCP server access controls
- Verify your Voka AI plan supports custom MCP servers
Security Best Practices
Protecting API keys:
- ✅ Use environment variables (not hardcoded)
- ✅ Rotate keys regularly
- ✅ Use read-only access when possible
- ✅ Limit IP addresses if supported
Data privacy:
- ✅ Review what data the MCP server accesses
- ✅ Only install trusted MCP servers
- ✅ Check the server's privacy policy
- ✅ Use separate API keys per environment (dev/prod)
Monitoring:
- ✅ Monitor MCP server logs
- ✅ Set up alerts for errors
- ✅ Review usage patterns regularly
- ✅ Audit data access periodically
Removing MCP Servers
To remove an installed MCP server:
- Go to AI Assistants → Edit assistant
- Agent tab → MCP Servers
- Find the server you want to remove
- Click Delete or Remove
- Confirm removal
Note: Removing an MCP server will break any assistant functionality that depends on it. Test thoroughly after removal.
Advanced: Multiple MCP Servers
Your assistant can use multiple MCP servers simultaneously:
Example configuration:
- Voka AI MCP (mcp.vokaai.com) - Acuity, Square, Jobber
- Custom Database MCP - Internal customer data
- Email MCP - Send confirmation emails
The assistant will automatically route queries to the appropriate server based on the request.