💻 Claude Code
Four features that turn Claude from a chatbot you paste into to a development workflow that runs your stack.
Claude Code is Anthropic's CLI and IDE integration — it reads your codebase, runs commands, and works across your tools. Most people use 10% of it. These are the four features that unlock the rest.
CLAUDE.md
Give Claude a permanent memory for your project
A file you drop in your project root that Claude reads at the start of every session — your stack, your decisions, your preferences, your conventions. No more re-explaining yourself every time.
How to set it up
Create a file called CLAUDE.md in your project root. Write it like a briefing document: what the project does, what stack you use, what patterns to follow, what to avoid. Claude reads it automatically on every session start.
Example
# My Project ## Stack - Next.js 14 (App Router), TypeScript, Tailwind, Prisma + PostgreSQL - Tests: Vitest + Playwright ## Conventions - Components: PascalCase, co-located with their tests - API routes: always validate with Zod before touching the DB - Never use any — use unknown and narrow properly ## Current priorities Working on: checkout flow refactor (see /src/checkout/) Do not touch: legacy /api/v1/ routes — deprecation in Q3
Skills
Teach Claude a workflow once — run it forever
Skills are reusable, named workflows you define once. When you type /skill-name, Claude executes the exact same process every time — no drift, no re-explaining, consistent results.
How to set it up
Create a .md file in your .claude/commands/ directory. Name it whatever you want the slash command to be. Describe the workflow in plain English — what to check, what to do, in what order. Claude runs it on demand.
Example
# /review You are reviewing a PR for this project. 1. Check all changed files follow our naming conventions (see CLAUDE.md) 2. Flag any `any` types — suggest specific types instead 3. Verify new API routes have Zod validation 4. Check for missing error handling in async functions 5. Summarize findings as: critical (blocking) / suggested / nitpicks
MCP
Connect Claude to your entire stack
Model Context Protocol lets Claude connect to Slack, GitHub, Google Drive, Linear, Postgres, and hundreds of other tools — reading, writing, and acting across your stack from a single conversation.
How to set it up
Add MCP servers to your Claude config. Each server exposes tools that Claude can call. Once connected, Claude can read Slack threads, create GitHub issues, query your database, or update Linear tickets — all from the same session.
Example
// .claude/mcp_servers.json
{
"github": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-github"],
"env": { "GITHUB_TOKEN": "your_token" }
},
"postgres": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-postgres"],
"env": { "DATABASE_URL": "postgres://..." }
}
} Routines
Automated tasks that run while you sleep
Routines let you hand Claude a recurring job — daily standup summaries, scheduled code reviews, monitoring tasks, report generation — and it runs on schedule with no human involvement.
How to set it up
Define a routine with a name, a schedule (cron syntax), and instructions. Claude executes it on schedule, uses any connected MCP tools it needs, and delivers results wherever you specify — Slack, email, a file, a GitHub comment.
Example
# Daily Standup Routine Schedule: 0 9 * * 1-5 # 9am, Mon-Fri Tasks: 1. Pull last 24h of commits from GitHub (MCP) 2. Check any open PRs waiting for review 3. List today's Linear tickets assigned to the team 4. Post summary to #dev-standup in Slack (MCP) Format: bullet list per person, max 3 items each
Further Reading
The official Model Context Protocol site — spec, server directory, quickstart guides, and the full list of tools Claude Code can connect to out of the box.
181 templates, 271-question quiz, agentic workflow patterns, hooks, plugins, and a security threat database — the best community-built deep-dive.
Claude Code Ultimate Guide
The best community-built resource for everything else — hooks, plugins, advanced agentic workflows, security threat database, 181 production-ready templates, and a 271-question quiz to test your knowledge.