Skip to content

architecture-summary.md

Latest commit

 

History

History
125 lines (91 loc) · 3.97 KB

architecture-summary.md

File metadata and controls

125 lines (91 loc) · 3.97 KB

Architecture Summary

The Three Tools

Cursor IDE

An AI-powered code editor that uses rules/*.mdc files for context and guidelines. Rules guide how the AI codes - they're passive, providing standards and patterns. You invoke them with @rule-name or they apply automatically based on file patterns.

Cursor CLI

A command-line interface for CI/CD that uses the same rules/ as the IDE. This runs AI operations from the terminal, typically with commands like cursor --command. It ensures your automated fixes follow the same standards as your interactive coding.

Claude Code

A command-line AI coding assistant by Anthropic that uses .claude/commands/*.md files for executable workflows. Commands do things - they're active, running tests or deployments. You invoke them with /command-name in your terminal.

Rules vs Commands

Rules (rules/*.mdc)

Rules provide context and guidelines. They're passive - they guide AI decisions without executing anything. Cursor IDE and Cursor CLI both use them. Examples: python-coding-standards.mdc tells the AI how to write Python, git-commit-message.mdc defines commit formats, django-models.mdc explains model structure patterns.

Commands (.claude/commands/*.md)

Commands define executable workflows. They're active - they run tasks and show results. Only Claude Code uses them. Examples: python-test.md runs pytest, python-lint.md runs ruff, deploy.md handles deployment.

They Serve Different Purposes

You can't port rules to commands or use commands in Cursor. They're fundamentally different: rules guide how AI codes, commands execute workflows. Both are valuable, and they work together naturally - rules ensure quality, commands automate tasks.

Repository Structure

~/.ai_coding_config/
├── .cursor/
│   ├── rules/              # Context for Cursor (IDE + CLI)
│   │   ├── python/
│   │   ├── django/
│   │   └── ...
│   └── settings.json       # Cursor preferences
├── .claude/
│   ├── commands/           # Workflows for Claude Code
│   │   ├── python-test.md
│   │   ├── python-lint.md
│   │   └── ...
│   ├── agents/             # Agent definitions
│   │   ├── test-writer.md
│   │   └── ...
│   └── settings.json       # Claude Code preferences
├── .mcp/servers/           # MCP configs (both tools)
├── prompts/                # AI setup prompts
└── templates/              # Project templates

How Developers Use Both

Cursor IDE Workflow

1. Open Python file
2. python-coding-standards.mdc auto-applies
3. AI suggests code following standards
4. Use Cmd+K for AI edits
5. Use @git-commit-message for commits

Claude Code Workflow

1. Write Python code
2. /python-lint - runs ruff
3. /python-test - runs pytest
4. /python-format - formats code
5. Commands execute and show results

Cursor CLI Workflow (CI/CD)

# In GitHub Actions
cursor --rules rules/ --check src/
# Uses rules to guide AI fixes

What We're Building

For Cursor Users

Rules: Already have many covering Python, Django, git, and more. Settings: Best practice configurations for the IDE. CLI configs: Setups for CI/CD use.

For Claude Code Users

Commands: Python and TypeScript workflows for testing, linting, deployment. Agents: Specialized assistants for specific tasks. Settings: CLI preferences.

For Both

MCP Servers work with both tools. Prompts provide AI-guided setup. Templates offer project starters.

Key Points

Rules and commands are fundamentally different. Cursor IDE and Cursor CLI share rules/ while Claude Code has its own .claude/commands/. Agents can be referenced by both tools - they're markdown files with frontmatter. Never mention .cursorrules - it's deprecated.

See Also