MCP Forge

<img src="https://r2cdn.perplexity.ai/pplx-full-logo-primary-dark%402x.png" class="logo" width="120"/> # --- # Project Best Practices for MCP Development ## MCP Best Practices 1. **Standardized Tool Naming** - Use descriptive, lowercase names with underscores - Follow the pattern: `[service]_[action]_[object]` - Example: `github_create_repository`, `todoist_update_task` 2. **Parameter Validation** - Always validate input parameters before processing - Return clear error messages for invalid inputs - Include parameter type specifications in tool definitions 3. **Error Handling** - Implement proper try/catch blocks for all API calls - Return structured error objects with code and description - Log errors for debugging but sanitize sensitive information 4. **Authentication Management** - Use environment variables for API keys and tokens - Never hardcode credentials in server code - Implement token refresh mechanisms where applicable 5. **Performance Optimization** - Cache repetitive API calls when appropriate - Implement request throttling to avoid rate limits - Split complex operations into smaller tools for better reusability ## Smithery-Specific Rules 1. **Configuration Schema** - Define complete `configSchema` in `smithery.yaml` - Document all configuration options thoroughly - Mark required parameters appropriately 2. **Command Generation** - Use the correct environment variable mapping in `commandFunction` - Ensure proper escaping of special characters in config strings - Follow the Smithery startup protocol for proper initialization 3. **Server Structure** - Implement required MCP protocol methods - Use the latest `@modelcontextprotocol/sdk` version - Structure projects with clear separation of concerns 4. **Deployment Best Practices** - Include proper Dockerfile configuration - Specify all dependencies in package.json - Set appropriate NODE_ENV for production deployments 5. **Versioning** - Use semantic versioning for all MCP servers - Document breaking changes between versions - Provide upgrade paths for major version changes ## Cursor IDE Integration 1. **MCP Server Registration** - Use consistent naming in `mcp.json` - Follow the standard command structure: ```json "command": "npx", "args": [ "-y", "@smithery/cli@latest", "run", "[package-name]", "--config", "[escaped-json-config]" ] ``` 2. **Tool Discovery** - Refresh tools after configuration changes - Verify green dot status for enabled servers - Check console for connection errors 3. **Configuration Management** - Escape quotes correctly in config strings (use `\"` for quotes inside config) - Use environment-specific configurations for different environments - Document required API keys and tokens 4. **Troubleshooting** - Check "No resources available" warnings in Cursor IDE - Verify network connectivity to API endpoints - Monitor token usage and rate limits 5. **Performance** - Order servers by frequency of use for faster tool discovery - Configure timeout settings appropriately - Use local servers for development when possible --- # Cursor MCP Rules ## Default Servers (IN ORDER) ### MCP Server 1: Sequential Thinking **Tool Name:** sequentialthinking **Description:** Structured problem-solving with hypothesis management and session logging. **When to use:** Complex logical problems, debugging, and step-by-step analysis. ### MCP Server 2: Todoist **Tool Names:** - todoist_create_task - todoist_get_tasks - todoist_update_task - todoist_delete_task - todoist_complete_task **Description:** Task management integration. **When to use:** Project planning, task tracking, and deadline management. ### MCP Server 3: GitHub **Tool Names:** - create_or_update_file - search_repositories - create_repository - get_file_contents - push_files - create_issue - create_pull_request - fork_repository - create_branch - list_commits - list_issues - update_issue - add_issue_comment - search_code - search_issues - search_users - get_issue **Description:** GitHub integration for repository and code management. **When to use:** Code collaboration, version control, and repository management. ## Additional MCP Servers ### Brave Search **Tool Names:** - brave_web_search - brave_local_search **Description:** Web search integration. **When to use:** Research, information gathering, and discovery. ### Perplexity Deep Research **Tool Name:** deep_research **Description:** Advanced AI-powered research. **When to use:** In-depth research requiring sophisticated analysis of web content. ### Devalexandre MCP Servers **Tool Names:** - pyppeteer_navigation - take_screenshot - click_element - exit_server **Description:** Browser automation tools. **When to use:** Web scraping, testing, and UI automation. ### Neon DB **Tool Names:** - __node_version - list_projects - create_project - delete_project - prepare_database_migration - complete_database_migration - describe_project - describe_table_schema - get_database_tables - create_branch - run_sql - run_sql_transaction - describe_branch - delete_branch - get_connection_string - provision_neon_auth **Description:** Database management. **When to use:** Database operations, migrations, and SQL queries. ### FireCrawl MCP **Tool Names:** - firecrawl_scrape - firecrawl_map - firecrawl_crawl - firecrawl_deep_research - firecrawl_batch_scrape - firecrawl_check_batch_status - firecrawl_check_crawl_status - firecrawl_search - firecrawl_extract **Description:** Advanced web scraping. **When to use:** Content extraction, site mapping, and large-scale web crawling. ## MCP Server Selection Guidelines - Use Sequential Thinking for complex problem-solving and planning - Use GitHub for all code and repository-related operations - Use Brave Search for quick information lookup - Use Todoist for project and task management - Use Neon DB for database operations - Use FireCrawl for advanced web scraping needs - Use Perplexity for in-depth research requiring sophisticated analysis - Use Devalexandre tools for browser automation and screenshots ## Tool Execution Protocol 1. Identify the appropriate tool based on the task requirements 2. Use the full tool name when invoking (e.g., "github_create_repository") 3. Provide all required parameters in the correct format 4. Handle any errors or unexpected responses appropriately 5. Confirm successful tool execution before proceeding ## Usage Notes - Always try MCP tools before manual implementation of functionality - If a tool fails, verify server status before attempting alternatives - Some tools require authentication tokens which must be configured - Complex workflows may require chaining multiple tools together - Respect rate limits and usage quotas of underlying services <div style="text-align: center">⁂</div> [^1]: https://pplx-res.cloudinary.com/image/upload/v1742942047/user_uploads/ZyqsNIkkkuKueqU/image.jpg [^2]: https://pplx-res.cloudinary.com/image/upload/v1742942066/user_uploads/RpFOwYlSkfbzvbh/image.jpg [^3]: https://ppl-ai-file-upload.s3.amazonaws.com/web/direct-files/17146172/bca66741-9637-4a7f-b0b7-3e88bf3e9130/mcp.rules.md [^4]: https://pplx-res.cloudinary.com/image/upload/v1742944001/user_uploads/uThfdtgctaUXHgK/image.jpg