Working with Cursor

Overview

This skill provides complete, authoritative documentation for Cursor IDE directly from cursor.com. Instead of guessing about rule formats, plugin structures, MCP configuration, or feature capabilities, read the official docs stored in this skill's references directory.

When to Use

Use this skill when:

• Writing or debugging .cursor/rules/*.mdc files
• Creating or configuring Cursor plugins
• Setting up MCP servers in Cursor
• Writing or testing Cursor skills
• Configuring Cursor agents or commands
• Setting up hooks for Cursor events
• Understanding Cursor AI modes (Agent, Ask, Plan, Debug)
• Using @-mention context injection
• Configuring Cursor settings or extensions
• Troubleshooting Cursor issues

Quick Reference

Task	Read This File
Write or debug rules	rules.md
Set up MCP servers	mcp.md
Create a plugin	plugins.md
Write a skill	skills.md
Configure hooks	hooks.md
Configure subagents	subagents.md
Agent mode overview	agent-overview.md
CLI usage	cli-overview.md
CLI slash commands	cli-reference-slash-commands.md
CLI auth	cli-reference-authentication.md
GitHub integration	integrations-github.md
BugBot	bugbot.md
Cloud agent	cloud-agent.md
Security	security-agents.md
Models & pricing	models-and-pricing.md
Get started	get-started-quickstart.md

Documentation Organization

All documentation is stored as individual markdown files in references/. Use the Read tool to access specific documentation:

references/
├── rules.md                       # Rules (.mdc format, activation modes, globs)
├── skills.md                      # Skill authoring guide
├── plugins.md                     # Plugin development
├── mcp.md                         # Model Context Protocol integration
├── subagents.md                   # Subagent configurations
├── hooks.md                       # Event hooks
├── agent-overview.md              # Agent mode overview
├── agent-plan-mode.md             # Plan mode
├── agent-agents-window.md         # Agents window UI
├── agent-tools-terminal.md        # Terminal tool
├── agent-tools-search.md          # Search tool
├── agent-tools-canvas.md          # Canvas tool
├── cli-overview.md                # CLI reference
├── cli-reference-slash-commands.md # Slash commands
├── cli-reference-authentication.md # Auth
├── cli-reference-configuration.md  # Config
├── cloud-agent.md                 # Cloud agent overview
├── cloud-agent-setup.md           # Cloud agent setup
├── cloud-agent-best-practices.md  # Cloud agent best practices
├── integrations-github.md         # GitHub integration
├── security-agents.md             # Security
├── models-and-pricing.md          # Model selection and pricing
├── get-started-quickstart.md      # Quickstart guide
└── ...                            # Additional docs (teams, enterprise, SDK)

Key Concepts

Rules vs Skills vs Plugins

• Rules (.cursor/rules/*.mdc) — persistent AI guidance injected into context; team-shareable via git
• Skills — reusable multi-step capabilities invoked by the agent
• Plugins — distributable packages bundling rules, skills, commands, MCP servers, and hooks together

Rule Activation Modes

• alwaysApply: true — loaded in every request (keep under 200 words)
• globs set — auto-attached when matching files are in context
• description set, no globs — agent decides when to apply based on description
• Neither — manual invocation only (@rule-name in chat)

MCP in Cursor Cursor supports MCP tools, prompts, and roots. Configure globally in ~/.cursor/mcp.json or per-project in .cursor/mcp.json. Cursor supports up to 40 MCP tools at a time.

Cross-References

For plugin development workflows, use the developing-cursor-plugins skill.