# THE CODING TOOLS DEEP DIVE
**Every major AI coding tool, compared honestly — in plain English.**
From alchemy-ai.dev — free forever, no affiliate links, no sponsored rankings.
Last updated: July 2026. Prices change fast — verify on the official site before paying.

---

## START HERE: THE FOUR KINDS OF CODING TOOLS

Before comparing tools, you need to know there are four different *kinds* — and most "which tool is best?" arguments are really two people talking about different categories.

**1. Terminal / editor agents** — AI that works inside your coding setup and does the work *for* you: reads your files, writes code across many files, runs commands. (Claude Code, Cline, Aider)

**2. AI-native editors** — a full code editor (like VS Code) rebuilt around AI. (Cursor, Windsurf)

**3. Editor add-ons** — AI bolted onto the editor you already use, historically focused on autocomplete. (GitHub Copilot)

**4. Browser app builders** — websites where you type "build me an app" and get a working demo, no coding setup at all. (Lovable, Bolt.new, Replit, v0)

Quick self-sort:
- **Never coded, want to see something on screen today?** → Category 4.
- **Following this site's CLAUDE.md workflow?** → Category 1 (Claude Code) is the backbone.
- **Comfortable in an editor, want AI everywhere in it?** → Category 2.
- **Just want cheap autocomplete while you type?** → Category 3.

---

## THE QUICK ANSWER TABLE

| Tool | Type | Free Tier? | Paid Starts At | Best For |
|---|---|---|---|---|
| **Claude Code** | Agent (terminal + VS Code) | With Claude free (limited) | $20/mo via Claude Pro | Full builds from a CLAUDE.md |
| **Cursor** | AI-native editor | Yes | $20/mo Pro | Developers living in their editor |
| **Windsurf** | AI-native editor | Yes (generous) | $20/mo Pro | Cursor alternative, friendlier free tier |
| **GitHub Copilot** | Editor add-on + agent | Yes | $10/mo Pro | Cheapest capable paid option |
| **Cline** | Open-source agent (VS Code) | Free tool + pay per use | ~$5–20/mo in API usage | Control freaks (affectionate) |
| **Aider** | Open-source agent (terminal) | Free tool + pay per use | ~$5–20/mo in API usage | Terminal people, git lovers |
| **Lovable / Bolt.new / v0** | Browser builders | Yes (small) | ~$20–25/mo | Non-coders, fast demos |
| **Replit** | Browser builder + IDE | Yes | ~$25/mo | Learning + building in one place |

---

## CLAUDE CODE (Anthropic)

**What it is:** Claude, but instead of chatting in a browser you point it at your project folder. It reads your files, writes and edits code across your whole project, runs commands, and fixes its own errors. Available in the terminal, as a VS Code extension, and as a desktop app.

**What it costs:** Included with a Claude subscription. **Claude Pro ($20/month) covers most individual builders.** Max ($100–200/month) for heavy daily use. You can also pay per token through the API instead.

**THE GOOD**
- The best pure building agent available. Give it a filled-out CLAUDE.md and it plans and executes multi-file builds step by step.
- One $20 subscription covers chat on claude.ai AND Claude Code. Two tools, one bill.
- Handles the boring parts — creating files, wiring pages together, fixing the error it just caused — without you touching each file.
- This entire site was built with it. That's the proof of concept.

**THE BAD**
- Usage limits. A long build session on Pro can hit the cap; then you wait or upgrade.
- It's an agent, so it will occasionally do something you didn't ask for. Review what it changed — don't just click accept.
- Less hand-holding than browser builders. You need VS Code (or a terminal) set up first — see the VS Code Setup page.

**THE UGLY**
- If you use the pay-per-token API route instead of a subscription, costs can silently pile up. For beginners: use the subscription, not the API.
- When it gets stuck in a circular error loop, it will confidently keep trying the same broken idea. That's why the workflow says: take the error to ChatGPT, bring the answer back.

**Who it's for:** Anyone following this site's workflow. This is the build tool the CLAUDE.md template was designed for.

---

## CURSOR

**What it is:** A code editor (a modified VS Code) with AI built into every corner — autocomplete, chat with your codebase, and an agent mode that edits multiple files.

**What it costs:** Free tier; **Pro $20/month**; Pro+ $60/month and Ultra $200/month for heavy users; Business $40/user/month.

**THE GOOD**
- The smoothest "AI everywhere" editor experience. Tab-autocomplete feels like mind reading on repetitive code.
- Understands your whole codebase — ask "where do we handle login?" and it knows.
- Lets you pick between different AI brains (Claude, GPT, Gemini) inside one tool.

**THE BAD**
- It replaces your editor. If you have years of VS Code muscle memory and extensions, migration is friction (though it's VS Code-based, so most things carry over).
- The $20 Pro tier has usage limits that power users blow through fast — that's what the $60 and $200 tiers exist for.

**THE UGLY**
- Pricing and usage-limit changes have repeatedly angered its own community — plans have been redefined mid-subscription before. Read current terms, not last year's blog posts.
- Costs stack: many pros end up paying for Cursor AND a Claude plan.

**Who it's for:** Developers who code daily and want AI woven into the editing experience itself. Overkill if Claude Code already does your building.

---

## WINDSURF

**What it is:** Cursor's main rival — another AI-native editor, known for its "Cascade" agent and a friendlier free tier.

**What it costs:** Generous free tier (unlimited autocomplete); **Pro $20/month** (raised from $15 in May 2026); Max tier around $200/month.

**THE GOOD**
- The best free tier among AI editors — real daily use without paying.
- Cascade agent mode is genuinely good at multi-file changes.
- Feels a bit simpler and less overwhelming than Cursor for newcomers.

**THE BAD**
- Smaller community than Cursor — fewer tutorials, fewer answered questions when you're stuck.
- Feature-for-feature it usually trails Cursor by a few months.

**THE UGLY**
- The company has been through acquisition drama (talent absorbed by bigger players in 2025), which left real questions about its long-term direction. It's alive and improving, but bet accordingly.

**Who it's for:** People who want the AI-editor experience without paying on day one, or who tried Cursor and didn't click with it.

---

## GITHUB COPILOT (Microsoft/GitHub)

**What it is:** The original AI coding assistant, living inside VS Code and other editors. Started as autocomplete; now includes chat and an agent mode. Owned by Microsoft, wired deep into GitHub.

**What it costs:** **Free tier** (limited completions and chats); **Pro $10/month** — the cheapest capable paid plan in this list, and it includes access to top-tier models including Claude. Heads up: GitHub is shifting toward usage-based billing (metered "premium requests"), so heavy agent use can exceed the flat fee.

**THE GOOD**
- $10/month is the best entry price in AI coding, and the free tier is a real trial.
- Zero-friction setup if you already use VS Code and GitHub — it's one extension.
- You can pick which AI brain answers (including Claude models) inside Copilot.

**THE BAD**
- Its agent mode trails Claude Code and Cursor for big multi-file builds. It shines at autocomplete and small tasks, not driving a whole project.
- Spread across so many Microsoft surfaces that quality varies by where you use it.

**THE UGLY**
- The pricing model is drifting from "flat $10" to "flat $10 plus metered charges." Watch your premium-request count or the bill can surprise you.
- Confusion trap: **GitHub Copilot (coding) and Microsoft Copilot (Office) are different products with the same name.** Articles mix them up constantly.

**Who it's for:** Budget-conscious coders, students (it's free for verified students!), and anyone who mostly wants smart autocomplete rather than a full building agent.

---

## CLINE & AIDER (Open Source)

**What they are:** Free, open-source agents that do what Claude Code does — but you "bring your own key" (BYOK): you sign up for API access with an AI provider (Anthropic, OpenAI, DeepSeek, etc.), get a key (a password for pay-per-use AI), and plug it in. **Cline** runs in VS Code; **Aider** runs in the terminal and is loved for its tight git integration.

**What they cost:** The tools are free. You pay per token to whichever AI provider you connect — light use runs $5–20/month, and pairing them with a cheap model (like DeepSeek) can drop that to pocket change.

**THE GOOD**
- Total control: pick any AI brain, see exactly what every request costs, swap providers anytime.
- No subscription, no lock-in. Light-month = tiny bill.
- Open source: the community can inspect the code, and improvements ship fast.

**THE BAD**
- Setup involves API keys, provider accounts, and config — a real step up in friction from "install extension, sign in."
- No one to call when it breaks. Support is GitHub issues and Discord.

**THE UGLY**
- **An unwatched agent on a pay-per-token key can burn real money fast.** One runaway loop on an expensive model can cost more than a month of Claude Pro. Set spending limits in your provider dashboard on day one — not optional.
- API keys are secrets. Leak one in public code and someone else spends your money.

**Who they're for:** Tinkerers and developers who want maximum control and minimum subscription. Not beginners — do the subscription route first, graduate to this later.

---

## THE BROWSER BUILDERS — Lovable, Bolt.new, v0, Replit

**What they are:** Websites where you describe an app in plain English and watch it get built in your browser. No installs, no setup. **Lovable** and **Bolt.new** target non-coders building web apps; **v0** (by Vercel) specializes in user-interface designs; **Replit** is a full online coding environment with an AI agent, great for learning.

**What they cost:** All have free tiers big enough to try; paid plans cluster around $20–25/month, usually metered by usage credits.

**THE GOOD**
- Fastest possible "I have an idea → I can see it working" path. Minutes, not days.
- Nothing to install. Works from a Chromebook, a library computer, anywhere.
- Great for validating an idea before committing to a real build.

**THE BAD**
- The demo ceiling is real: first 80% feels magical, the last 20% (login, payments, edge cases) gets hard fast — and you can't easily drop down to fix things by hand like you can in VS Code.
- Credit-based pricing means a stubborn bug can eat your monthly allowance while you re-prompt.

**THE UGLY**
- Vendor lock-in varies. Some make it easy to export your code and leave; some make it painful. Check *before* you build something you care about.
- "You'll never need to code" is marketing. Every serious app built on these eventually needs someone who can read the code.

**Who they're for:** Absolute beginners, idea validators, and quick demos. A great on-ramp — and many people outgrow them into the CLAUDE.md workflow, which is exactly what this site is for.

---

## THE BOTTOM LINE

**The setup this site teaches:** VS Code (free) + Claude Code via Claude Pro ($20/month) + ChatGPT free for debugging. Total: $20/month, and it can build and ship real applications.

**The budget path:** GitHub Copilot Pro ($10/month) or Windsurf free — less powerful for full builds, but real.

**The $0 path:** Windsurf free tier + Claude free + ChatGPT free. Slower, capped, but genuinely workable for learning.

**The pro path:** Most professionals stack — an AI editor (Cursor/Windsurf) for daily editing plus Claude Code for the heavy builds. Two subscriptions, and they earn their keep.

Whichever you pick: **start with one tool, finish one small project with it, then decide.** Tool-hopping is the most expensive hobby in AI.

---

*From alchemy-ai.dev — a free AI education platform by Stacked Alchemist LLC. No course to sell. No paywall.*