Coding tools

Start with

Learn

• Does my AI subscription pay for API use too?

  Usually not. ChatGPT and Claude plans are billed separately from their APIs. The exceptions are the companies' own apps, and one setting can quietly switch Claude Code to the API.

• When should Claude Code use a subagent?

  A subagent is a helper Claude Code sends off to do one focused job and report back. When that saves you a cluttered session, and when it only burns more of your allowance.

• Skills: teach your coding agent something once

  A Claude Code skill is a short written procedure the agent reads when a job matches. Write one for anything you have explained twice.

• You may already pay for a coding agent

  If you pay for ChatGPT, Codex is included: a coding agent that works on your real files. Most subscribers never open it. Try it in five minutes.

• Compact or restart? Keeping Claude Code sharp

  Long coding sessions go stale for the same reason long chats do. Claude Code gives you two ways out, and picking the right one makes or breaks the afternoon.

Guides

• Claude Code permissions: what each mode actually lets it do

  Ask every time, accept edits, or let it run: three Claude Code settings, three very different afternoons. Plan mode vs auto mode, which to use when, and the safe rule.

• Hit your Claude limit? The two limits and what to do next

  Claude's paid plans have a five-hour limit and a weekly one. When each resets, how to tell which you hit, Pro vs Max, and how to make the same work use less.

• Undo what Claude Code just did: rewind and checkpoints

  Claude Code saves a checkpoint before every message you send. Press Esc twice or type /rewind to put your files back, and know the one kind of change it can't undo.

• CLAUDE.md: the note your coding agent reads first

  Claude Code reads one file before every task. Most people leave it empty or fill it with junk. What goes in, what stays out, and a prompt that writes it.

• Before you add an MCP server, check what it can do

  Connecting an MCP server hands your AI new powers, sometimes including sending and deleting. A ten-minute check of who made it, what its tools do, and how to take it away again.

• Make Claude Code check its own edits with hooks

  A hook is a command Claude Code runs by itself at a set moment, like after every edit. Set up one that tidies each file it touches, check it works, and know how to turn it off.

• Write your first Claude Code skill for a repeated job

  Stop pasting the same long prompt. Save the procedure once as a skill, run it by name, check it does what you wrote, and delete it when it stops helping.

• Plan mode: make Claude Code think before it touches anything

  For anything bigger than a one-line fix, make Claude Code plan first and change nothing. Read the plan, fix the wrong assumption, then let it go.

• Git in four commands (so the agent can never wreck your project)

  You do not need to understand Git. You need four commands that make every change an AI agent makes reversible: save point, check, undo, go back.

• One task at a time (the habit that stops the mess)

  Ask AI for five things in one message and you get a muddle you cannot check. Ask for one, check it, then the next. Slower per message, faster per job.

• Send one message at 7am so your AI limit resets at lunch

  Claude and Codex limit use in windows of a few hours that start with your first message. Start the clock early and get two full windows in a working day.

• Get a website design you like (without being a designer)

  Stop describing the page you want. Ask AI to draw several versions, pick one, then hand that picture to a coding AI and let it build the real thing.

Compare

• Claude Code vs Codex: which coding agent?

  Both work in your terminal on your real files, and both come with a plan you may already pay for. I run both. Where each one wins, and how to decide.

Tools

• Claude: the best writer of the big assistants

  Anthropic's assistant. Less padding, more willing to say it is not sure, and the paid plan includes Claude Code, the agent that built this site. No pictures.

• GitHub Copilot: AI inside the editor most developers already use

  Code suggestions, chat and an agent mode inside VS Code and the other big editors. The default choice at work, and a fair one if you are already in that world.

• Lovable: describe an app, get a working one

  Type what you want in plain English and get a working web app with a login, hosted, in minutes. Astonishing for a first version; the hard part comes after.

• Windsurf: the other agent-in-an-editor

  An editor with an AI agent built in, the same shape as Cursor. Cheaper plans, a slightly different feel, and the same rule: watch the edits till you trust it.

• Codex: OpenAI's coding agent in your terminal

  OpenAI's coding agent. It works in your terminal on your real project files, and the models behind it can be pointed at other tools that need one.

• Cursor: an AI coding agent inside your editor

  A code editor with an AI agent built into it. Same idea as a terminal agent, except you watch the changes land in the file as they happen.

• Claude Code: what it is and whether you need it

  An AI coding assistant that works in your terminal and inside your real project files, rather than in a chat window you copy and paste from.

Setups

• My current AI setup

  The exact set of tools I use day to day, what each one is for, and what changed when I stopped writing this page from memory and actually checked.