_{🛠️ Methodologies · 🧠 Memory systems · 🔌 MCP servers · 📜 Templates · 🤖 Skills · 🎬 Videos · 📚 Reading}

A curated list of practical resources for shipping production software with Claude Code. Not chatbots, not toys. Real methodologies, real MCP servers, real templates you can copy-paste into your project today.

Contents

• 📜 Methodologies
• 🧠 Memory + Knowledge systems
• 🔌 MCP servers
• 📂 Awesome MCP catalogs
• 🎨 Skills
• 🤖 Agents
• 📜 CLAUDE.md / HANDOFF.md templates
• 🛡 Safety guards
• 📚 Reading
• 🎬 Videos & talks
• 🌐 Official Anthropic resources
• Contributing

---

📜 Methodologies

How to actually organize your project so Claude can ship without re-explaining the world each session.

• agnis-workflow — Persistent memory (4 typed file types), CTO mode, iteration loops, two-boots research agents. Used on 5+ shipped products.
• Andrej Karpathy's wiki idea — Original tweet on .md files as long-term AI memory. The seed many later workflows extend.
• Boris Cherny — iteration loops — Measure → Change → Measure → Deploy → Record. Adapted by most production AI workflows.

🧠 Memory + Knowledge systems

Long-term memory that persists across sessions. Filesystem-as-database approach.

• agnis-workflow/templates/memory — Templates for user_profile.md, feedback_*.md, project_*.md, reference_*.md with frontmatter typing.
• Anthropic memory cookbook — Official patterns for context engineering and memory.

🔌 MCP servers

Model Context Protocol lets Claude (and other LLM hosts) call external tools as first-class capabilities. These are individual servers worth wiring into your stack.

Russian-market platforms

• SergeyKrin9/mcp-servers — Yandex.Direct, Yandex.Webmaster, Yandex.Metrika, GSC, YouTube, VK, Telegram, Click.ru. The only unified RU-market MCP bundle.

General-purpose

• modelcontextprotocol/servers — Anthropic's reference servers (Filesystem, Git, GitHub, Postgres, …).
• upstash/context7 — Docs lookup MCP. The most popular single-server.
• microsoft/playwright-mcp — Browser automation via Playwright. Essential for web research and testing.
• microsoft/mcp — Microsoft's collection (Office, Azure, …).

📂 Awesome MCP catalogs

Where to discover new MCP servers and submit your own.

• punkpeye/awesome-mcp-servers — The de-facto largest catalog. ~88k ⭐.
• modelcontextprotocol/servers — Official Anthropic reference.
• Official MCP Registry — Anthropic + GitHub + Microsoft registry. Requires server.json + reverse-DNS namespace + public npm/PyPI/Docker package.
• mcp.directory · mcp.so · glama.ai/mcp/servers · pulsemcp.com · mcpmarket.com — Submission directories. Each is a free listing.
• wong2/awesome-mcp-servers — Second largest catalog (mcpservers.org).
• tolkonepiu/best-of-mcp-servers — Star-ranked list with medals.
• appcypher/awesome-mcp-servers — Alternative awesome list.

🎨 Skills

Named workflows the AI invokes for repeated tasks. /seo-audit, /code-review, /copy-edit, …

• davidteren/claude-skills — Community-curated skills.
• anthropics/anthropic-skills — Official skills (xlsx, pdf, docx, skill-creator, …).
• hesreallyhim/awesome-claude-code — Larger collection that includes skills as one section.

🤖 Agents

Long-running tools that handle isolated work — codebase exploration, parallel research, multi-agent decision-making.

• All-Hands-AI/OpenHands — Open-source autonomous agent. SWEBench top performer.
• cline/cline — VS Code agent.
• AntonOsika/gpt-engineer — Specify a task, AI ships the code. One of the originals.
• continuedev/continue — VS Code + JetBrains IDE assistant with agent mode.
• e2b-dev/awesome-ai-agents — Curated list of autonomous agents.

📜 CLAUDE.md / HANDOFF.md templates

• agnis-workflow/templates — Battle-tested templates: CLAUDE.md, HANDOFF.md, MEMORY.md, first-session prompt.
• Anthropic — Claude Code best practices — Official guidance on CLAUDE.md structure.

🛡 Safety guards

Scripts and patterns that stop you from breaking production.

• agnis-workflow/examples/pre-deploy-check.sh — Generic pre-deploy guard, configurable danger-file list. Built after a real Firestore.rules incident.
• detect-secrets — Pre-commit hook to stop secret leaks.
• gitleaks — Alternative secret scanner with better default rules.

📚 Reading

Foundational reading for the AI-as-runtime mindset.

• Anthropic — Engineering blog — Their team writes about how they use Claude internally.
• Anthropic — Building effective agents — Patterns for agentic systems.
• The Bitter Lesson (Rich Sutton) — Why methods that scale with compute beat clever methods.
• Andrej Karpathy — Software 2.0 — Why prompt + model is becoming the new compiler.
• Simon Willison's blog — Sharpest day-to-day commentary on LLM tooling.

🎬 Videos & talks

• Anthropic — Claude Code overview — Official videos.
• Andrej Karpathy — Let's build GPT — From-scratch transformer that demystifies what the AI actually does.
• Lex Fridman + Andrej Karpathy — 5-hour deep dive. Worth it for the agent design intuition.

🌐 Official Anthropic resources

• Claude Code — The CLI tool.
• Claude Code docs — Official documentation.
• claude-code GitHub — Open issues and feature requests.
• Model Context Protocol — The MCP spec.
• Anthropic cookbook — Official patterns for the Claude API.
• claude-cookbooks — Production-ready notebooks.

---

Contributing

PRs welcome. Two rules:

1. It must work. No vaporware, no "coming soon" links. Items either ship today or move to a future Wishlist section.
2. It must teach. Lists of names with no description are noise. Every item gets one sentence on what it gives you that you can't get elsewhere.

PR format:

- **[Name](https://link)** — One sentence on what makes this worth reading/using.

License

To the extent possible under law, Sergey Krinitsyn has waived all copyright and related rights to this work.

---

Maintained by Sergey Krinitsyn · Email · GitHub

_{If this list saved you research time — drop a ⭐. It's the only signal I track.}