---
title: "AI Agents: Claude Code, Skills, MCP and Protocols"
description: "How AI agents work and how I use them: the agent loop, Claude Code skills, MCP tools, the A2A and ARD protocols, and cleaning up the worktrees agents leave."
canonical: "https://www.trevorlasn.com/topics/ai-agents"
author: "Trevor I. Lasn"
last-updated: "2026-11-11"
---

# AI Agents: Claude Code, Skills, MCP and Protocols

How AI agents work and how I use them: the agent loop, Claude Code skills, MCP tools, the A2A and ARD protocols, and cleaning up the worktrees agents leave.

14 current articles on AI agents, newest first. See [every topic](https://www.trevorlasn.com/topics.md).

## 2026

- [DeepSeek Harness Keeps Uploading Your Sessions After You Turn Telemetry Off](https://www.trevorlasn.com/blog/deepseek-harness-keeps-uploading-your-sessions-after-you-turn-telemetry-off.md): DeepSeek Harness sends your session log, plugin list and device id to DeepSeek by default. The documented opt-out stops one of four uploads. (2026-11-11)
- [Anthropic Has No Moat: The Bear Case Before the IPO](https://www.trevorlasn.com/blog/anthropic-has-no-moat.md): Anthropic's IPO values it near $1.4 trillion. What a moat is, why the Claude models don't have one, whether a rival can crush it, and why I wouldn't short it. (2026-11-10)
- [residue-cli: Reclaim your storage from Git worktrees, automatically and safely.](https://www.trevorlasn.com/blog/residue-cli-reclaim-storage-from-git-worktrees.md): residue-cli removes a Git worktree only when nothing in it would be lost. How its 15 checks decide, how it deletes and restores, and residue-ui for the Mac. (2026-11-09)
- [Agent Git worktrees eat your disk; residue-cli clears safe ones](https://www.trevorlasn.com/blog/residue-git-worktree-cleanup.md): residue-cli is an open-source Rust CLI that finds the Git worktrees coding agents leave on disk and removes one only when nothing in it would be lost. (2026-11-06)
- [Agentic Resource Discovery: What I Added to 0xinsider](https://www.trevorlasn.com/blog/agentic-resource-discovery-ard.md): ARD is a new proposal from Google, Microsoft and Hugging Face for how agents find tools. What's in it, what 0xinsider publishes, and what a registry returns. (2026-09-23)
- [Google Search Console MCP Server: Connect It to Claude Code](https://www.trevorlasn.com/blog/google-search-console-mcp-claude-code.md): Connect Google Search Console to Claude Code with the mcp-gsc MCP server: setup with uvx, the tools it adds, and the indexing and CTR problems it found. (2026-02-15)
- [AI Agents Explained](https://www.trevorlasn.com/blog/ai-agents-explained.md): What AI agents are, how the agent loop works, and why they're different from chatbots. (2026-01-30)
- [Building Custom MCP Servers with Next.js and mcp-handler](https://www.trevorlasn.com/blog/building-custom-mcp-servers-with-nextjs-and-mcp-handler.md): Build an MCP server in a Next.js route with mcp-handler and createMcpHandler, then connect it to Claude Code, Gemini CLI, Cursor and other MCP clients. (2026-01-28)

## 2025

- [Chrome DevTools MCP: Let Your AI Agent Debug Your App](https://www.trevorlasn.com/blog/chrome-devtools-mcp.md): Give your AI agent access to your running application. They can see errors, inspect the network tab, check the DOM, and debug issues while you work. (2025-11-10)
- [Claude Code Superpowers: Add Skills That Plan Before Coding](https://www.trevorlasn.com/blog/superpowers-claude-code-skills.md): Superpowers is a plugin for Claude Code that enforces planning, TDD, and debugging workflows. Setup guide with real examples. (2025-11-04)
- [Next.js DevTools MCP: Connect Agents to Your Dev Server](https://www.trevorlasn.com/blog/next-js-devtools-mcp.md): The Next.js DevTools MCP connects Claude and Cursor to your running dev server. I use it every day. (2025-11-02)
- [Context7 MCP: Stop LLM Hallucinations with Live Docs](https://www.trevorlasn.com/blog/context7-mcp.md): Context7 MCP server pulls version-specific library docs into Claude Code, Cursor, and other AI editors. Setup guide and how it works. (2025-11-01)
- [The Sequential Thinking MCP Server](https://www.trevorlasn.com/blog/anthropic-sequential-thinking-mcp.md): The sequential thinking MCP server has an AI agent reason one numbered thought at a time, revising or branching as it goes. What it does and when it helps. (2025-09-07)
- [Agent2Agent (A2A) Protocol: How AI Agents Talk to Each Other](https://www.trevorlasn.com/blog/agent-2-agent-protocol-a2a.md): An exploration of Google's new open protocol that enables different AI systems to exchange information and collaborate (2025-04-13)
