Overview
Follow these best practices to build reliable, efficient, and user-friendly integrations.Tool Design
Clear Naming
Use descriptive tool names that explain what they do:Detailed Descriptions
Write clear descriptions that help the AI understand when to use each tool:Parameter Documentation
Document all parameters with types and descriptions:Authentication
Secure Credentials
Never expose credentials in code or logs:Rotate Regularly
Update API keys and tokens periodically:- API Keys: Rotate every 90 days
- OAuth Tokens: Refresh before expiration
- Passwords: Change quarterly
Minimal Permissions
Request only the permissions your integration needs:Error Handling
Return Clear Errors
Provide helpful error messages:Handle Rate Limits
Respect API rate limits:Validate Inputs
Check all inputs before making API calls:Performance
Fast Responses
Keep tool responses under 5 seconds:Cache When Possible
Cache frequently accessed data:Batch Operations
Batch requests when possible:Testing
Test All Tools
Test every tool before deploying:Handle Edge Cases
Test error scenarios:Monitor Production
Check logs regularly: Dashboard -> MCP Servers -> Your Server -> Logs- Monitor error rates
- Check response times
- Track authentication issues
Documentation
Document Use Cases
Explain when to use your integration:Provide Examples
Include practical examples:Keep Updated
Update documentation when you change tools:- New parameters
- Changed behavior
- Deprecated features
Monitoring
Track Usage
Monitor tool usage: Agent -> Insights -> Tool Analytics- Most used tools
- Success rates
- Response times
Set Alerts
Get notified of issues:- High error rate: >5% failures
- Slow responses: >5 second average
- Auth failures: Any authentication errors
Review Regularly
Weekly review checklist:- Check error logs
- Review response times
- Verify rate limits not exceeded
- Update credentials if needed
- Check for API changes