# Cursor Ide > Work effectively with Cursor IDE - the AI-first code editor. Understand Cursor's agent mode, rules system, MCP integration, and how to create custom skills and rules for optimal AI-assisted development. Use when this capability is needed. - Skill: `tomevault-io/cursor-ide` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/cursor-ide` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/cursor-ide/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/cursor-ide --- # Cursor IDE — AI-First Development Guide [Cursor](https://cursor.com) is an AI-first code editor built on VS Code. This skill covers how to work effectively with Cursor's AI features, create custom rules, and integrate with Hermes Agent. ## Cursor Architecture ### Key Components | Component | Purpose | |-----------|---------| | **Agent Mode** | Autonomous coding with file access, terminal, and tool use | | **Ask Mode** | Read-only exploration and Q&A (no file modifications) | | **Rules System** | Persistent AI instructions (`.cursor/rules/`, `AGENTS.md`) | | **MCP Servers** | External tool integration via Model Context Protocol | | **Composer** | Multi-file editing and project-wide changes | | **Chat** | Conversational AI assistance | ### Cursor vs Claude Code | Feature | Cursor | Claude Code | |---------|--------|-------------| | Base | VS Code fork | Standalone CLI | | Interface | GUI IDE | Terminal TUI | | Rules | `.cursor/rules/*.md`, `AGENTS.md` | `CLAUDE.md`, `.claude/rules/` | | MCP | Built-in MCP support | `claude mcp add` | | Modes | Agent, Ask, Edit | Interactive, Print | | Best For | Visual development | CLI automation | ## Rules System Cursor uses a hierarchical rules system to provide persistent context to the AI. ### Rule Locations | Location | Scope | Purpose | |----------|-------|---------| | `.cursor/rules/*.md` | Project | Project-specific rules (git-tracked) | | `AGENTS.md` | Project root | Primary project instructions | | `~/.cursor/rules/*.md` | Global | Personal rules across all projects | | Cursor Settings | Global | UI-configured rules | ### Creating Effective Rules #### File: `.cursor/rules/coding-standards.md` ```markdown # Coding Standards ## TypeScript - Use strict mode (`"strict": true` in tsconfig) - Prefer `interface` over `type` for object shapes - Use `const` assertions for literal types - Always specify return types on functions ## React - Use functional components with hooks - Prefer named exports over default exports - Use React.FC sparingly (prefer explicit props typing) - Keep components under 200 lines ## Testing - Write tests alongside implementation - Use descriptive test names: `it('should X when Y')` - Mock external dependencies, not internal modules ``` #### File: `AGENTS.md` ```markdown # Project: My App ## Overview A Next.js 14 application with TypeScript, Tailwind CSS, and Prisma ORM. ## Architecture - `/app` - Next.js App Router pages and layouts - `/components` - Reusable React components - `/lib` - Utility functions and shared logic - `/prisma` - Database schema and migrations ## Commands - `npm run dev` - Start development server - `npm run build` - Production build - `npm run test` - Run Jest tests - `npm run lint` - ESLint + Prettier ## Conventions - Use kebab-case for file names - Use PascalCase for components - Use camelCase for functions and variables - Prefix hooks with `use` - Prefix context providers with `Provider` ## Database - Always create migrations for schema changes - Use transactions for multi-table operations - Index foreign keys and frequently queried fields ``` ### Rule Best Practices 1. **Be Specific**: "Use 2-space indentation" not "Format code nicely" 2. **Include Examples**: Show the pattern you want, not just describe it 3. **Organize by Topic**: Separate rules into focused files 4. **Update Regularly**: Rules should evolve with the project 5. **Don't Duplicate**: Reference other rules instead of copying ## Agent Mode Agent Mode gives Cursor autonomous capabilities to: - Read and write files - Execute terminal commands - Search the codebase - Make multi-file changes ### Enabling Agent Mode 1. Open Cursor Settings (Cmd/Ctrl + ,) 2. Navigate to "Features" → "Agent" 3. Enable "Agent Mode" 4. Configure allowed tools and permissions ### Agent Mode Prompts Effective prompts for Agent Mode: ``` # Good: Specific and actionable "Add error handling to all API routes in /app/api/. Wrap handlers in try-catch, log errors with console.error, and return appropriate HTTP status codes." # Good: Multi-step with clear outcome "Refactor the authentication system: 1. Extract auth logic from pages into /lib/auth.ts 2. Create useAuth hook for client-side auth state 3. Add middleware for protected routes 4. Update all pages to use the new system" # Bad: Vague "Make the code better" # Bad: Too broad "Rewrite everything in TypeScript" ``` ### Agent Mode Safety Configure permissions in Cursor Settings: ```json { "cursor.agent.allowedTools": ["read", "write", "search"], "cursor.agent.requireConfirmation": true, "cursor.agent.maxFilesPerOperation": 10 } ``` ## MCP Integration Cursor supports Model Context Protocol (MCP) for external tool integration. ### Configuring MCP Servers #### File: `.cursor/mcp.json` ```json { "mcpServers": { "github": { "command": "npx", "args": ["@modelcontextprotocol/server-github"], "env": { "GITHUB_TOKEN": "${env:GITHUB_TOKEN}" } }, "postgres": { "command": "npx", "args": ["@anthropic-ai/server-postgres"], "env": { "DATABASE_URL": "${env:DATABASE_URL}" } }, "filesystem": { "command": "npx", "args": ["@anthropic-ai/server-filesystem", "--root", "${workspaceFolder}"] } } } ``` ### Using MCP Tools Once configured, MCP tools appear in Cursor's tool palette: ``` @github:issue://123 # Reference a GitHub issue @postgres:query # Execute a database query @filesystem:read # Read files via MCP ``` ### Popular MCP Servers | Server | Purpose | Install | |--------|---------|---------| | `@modelcontextprotocol/server-github` | GitHub integration | `npx` | | `@anthropic-ai/server-postgres` | PostgreSQL queries | `npx` | | `@anthropic-ai/server-filesystem` | File operations | `npx` | | `@anthropic-ai/server-puppeteer` | Browser automation | `npx` | | `@anthropic-ai/server-memory` | Persistent memory | `npx` | ## Creating Cursor Skills Cursor skills are markdown files that provide specialized knowledge. ### Skill Location ``` ~/.cursor/skills/ ├── my-skill/ │ └── SKILL.md └── another-skill/ ├── SKILL.md └── templates/ └── example.md ``` ### Skill Structure ```markdown --- name: my-skill description: Brief description of what this skill does version: 1.0.0 author: Your Name --- # Skill Name ## Overview What this skill helps with. ## When to Use - Scenario 1 - Scenario 2 ## Instructions Step-by-step guidance for the AI. ## Examples Concrete examples of usage. ## Templates Reusable patterns and boilerplate. ``` ### Skill Best Practices 1. **Single Responsibility**: One skill per domain 2. **Include Examples**: Show, don't just tell 3. **Provide Templates**: Reusable code snippets 4. **Document Triggers**: When should the AI use this skill? 5. **Version Control**: Track skill changes ## Keyboard Shortcuts ### Essential Shortcuts | Action | Mac | Windows/Linux | |--------|-----|---------------| | Open Chat | Cmd + L | Ctrl + L | | Open Composer | Cmd + I | Ctrl + I | | Toggle Agent Mode | Cmd + Shift + A | Ctrl + Shift + A | | Accept Suggestion | Tab | Tab | | Reject Suggestion | Esc | Esc | | Next Suggestion | Alt + ] | Alt + ] | | Previous Suggestion | Alt + [ | Alt + [ | | Inline Edit | Cmd + K | Ctrl + K | | Generate in Terminal | Cmd + K (in terminal) | Ctrl + K | ### Navigation | Action | Mac | Windows/Linux | |--------|-----|---------------| | Go to File | Cmd + P | Ctrl + P | | Go to Symbol | Cmd + Shift + O | Ctrl + Shift + O | | Go to Definition | F12 | F12 | | Find References | Shift + F12 | Shift + F12 | | Search Workspace | Cmd + Shift + F | Ctrl + Shift + F | ## Cursor + Hermes Integration ### Using Hermes from Cursor When working in Cursor with Hermes Agent running: 1. **Hermes as MCP Server**: Configure Hermes as an MCP server for Cursor 2. **Shared Rules**: Sync rules between `.cursor/rules/` and Hermes skills 3. **Complementary Roles**: Use Cursor for visual editing, Hermes for automation ### Hermes MCP Configuration ```json { "mcpServers": { "hermes": { "command": "hermes", "args": ["mcp", "serve"], "env": { "HERMES_HOME": "${env:HOME}/.hermes" } } } } ``` ### Workflow: Cursor + Hermes ``` 1. Use Cursor for: - Visual code editing - Interactive debugging - Real-time AI suggestions - Multi-file refactoring 2. Use Hermes for: - Background automation - CLI-based workflows - Scheduled tasks - Multi-platform messaging - Long-running operations ``` ## Troubleshooting ### Common Issues #### AI Not Following Rules 1. Check rule file location (`.cursor/rules/` or `AGENTS.md`) 2. Ensure rules are specific and actionable 3. Restart Cursor to reload rules 4. Check for conflicting rules #### MCP Server Not Connecting 1. Verify server is installed: `npx @server-name --version` 2. Check environment variables are set 3. Review Cursor's MCP logs (Output panel → MCP) 4. Test server manually: `npx @server-name` #### Agent Mode Not Working 1. Ensure Agent Mode is enabled in settings 2. Check file permissions 3. Verify workspace trust settings 4. Review allowed tools configuration ### Debug Mode Enable verbose logging: ```json { "cursor.debug.enabled": true, "cursor.debug.logLevel": "verbose" } ``` View logs: Help → Toggle Developer Tools → Console ## Best Practices ### For AI-Assisted Development 1. **Write Clear Commit Messages**: AI can learn from your history 2. **Use Descriptive Names**: Better names = better AI understanding 3. **Comment Complex Logic**: Help AI understand intent 4. **Keep Files Focused**: Smaller files = better context 5. **Update Rules Regularly**: Evolve rules with your project ### For Team Collaboration 1. **Share Rules via Git**: `.cursor/rules/` should be tracked 2. **Document Custom Skills**: Help teammates understand AI behavior 3. **Standardize MCP Config**: Use consistent server configurations 4. **Review AI Changes**: Always review before committing ### For Performance 1. **Limit Context Size**: Don't include entire codebase in prompts 2. **Use Specific References**: `@file.ts` instead of "the file" 3. **Break Large Tasks**: Smaller tasks = faster responses 4. **Cache MCP Results**: Avoid repeated expensive queries ## Resources - **Cursor Documentation**: https://docs.cursor.com - **Cursor Discord**: https://discord.gg/cursor - **MCP Specification**: https://modelcontextprotocol.io - **VS Code Docs**: https://code.visualstudio.com/docs (base editor) --- > Source: [mgkram4/hermes-local-ai-setup](https://github.com/mgkram4/hermes-local-ai-setup) — distributed by [TomeVault](https://tomevault.io).