What to Actually Buy

What I Actually Use

Here's my personal stack as of early 2026 and what I pay:

That's it. I've tried plenty of other tools and dropped them. More on that below.

Pick Your Budget

$0/mo — Just Evaluating

Enough to kick the tires. Not enough to form a real opinion — the limits are too tight to integrate into your daily workflow. If you're serious about evaluating, commit to at least one paid tier for a month.

$10/mo — The First Step

Get: GitHub Copilot Pro ($10/mo)

Unlimited inline completions. Always on, never in the way. You accept or reject with a keystroke — it doesn't change how you work, it just makes you faster. This is the lowest-risk entry point into AI-assisted coding.

$20/mo — The Real Start

Pick one:

• Claude Code Pro ($20/mo) — The best agentic tool. Multi-file refactors, debugging, PRs. Reads your entire codebase. This is where the real productivity unlock happens. Choose this if you're comfortable in the terminal.
• Cursor Pro ($20/mo) — All-in-one AI editor. Completions, inline editing with Cmd+K, chat with codebase context, agent mode. Choose this if you want everything inside one IDE.

Don't get both — they overlap significantly at this price point. My preference is Claude Code because it operates at a different layer than your editor and doesn't lock you into a specific IDE.

$30/mo — The Sweet Spot

This is the setup for most individual developers. Copilot handles flow-state completions (line-level), Claude Code handles complex work (task-level). They operate at different layers with zero conflict.

$45/mo — My Stack

Add CodeRabbit when you're shipping enough PRs that review time matters. It catches real issues — missed edge cases, architectural violations, inconsistent patterns — before any human sees your PR. If you're not sure whether to start with code review first, read the zero-risk adoption guide.

$60+/seat — Team

CodeRabbit pays for itself by cutting review cycles. Claude Code Teams gives shared billing and team-wide configurations.

Mixed teams (devs + designers + support):

Enterprise (10+): Contact vendors. Negotiate hard on SSO, audit logs, data retention, IP indemnity (Copilot Enterprise has it at $39/seat). Every vendor's enterprise tier is "contact us" pricing — they all negotiate.

Tools I've Tried and Dropped

Cursor — Good editor, but I'm already in VS Code with Copilot and using Claude Code in the terminal. Cursor's agent mode overlaps too much with Claude Code, and I don't want to maintain two sets of AI configurations. If you're starting fresh and want one tool, Cursor is fine. But Copilot + Claude Code is a better split because they operate at different layers with zero conflict.

Codex — OpenAI's autonomous background agent. The idea is great: assign it a task, it works in a sandbox, delivers a PR. In practice, the quality is inconsistent and you spend as much time reviewing and fixing its output as you would doing the task yourself. Maybe it'll get better. For now, Claude Code with explicit instructions produces more reliable results.

Lovable free tier — Fine for throwaway prototypes. Not something I'd pay for as a developer. Designers get much more value from it (see the designer guide).

The Full Lineup

Claude Code

The best agentic coding tool right now, and it's not close. It reads your entire codebase, executes commands, creates PRs, and actually understands your project's architecture when you give it a good CLAUDE.md.

My take: Pro is enough for most developers. The free tier is too limited to form a real opinion — if you're evaluating, commit to a month of Pro.

GitHub Copilot

The best value in AI coding. At $10/mo with unlimited completions, you're paying pennies per suggestion. It does one thing — inline completions — and does it well.

My take: Just get Pro. The chat is decent but I still prefer Claude Code for anything beyond simple questions.

Cursor

A VS Code fork with AI baked into every interaction. Tab completions, inline editing with Cmd+K, chat with codebase context, and an agent mode for multi-file tasks.

My take: Pick Cursor OR Copilot, not both. They overlap significantly for completions. Cursor wins if you want an all-in-one AI editor. Copilot wins if you like your current editor and want AI layered on top.

CodeRabbit

AI-powered automated code review on every PR. Configurable rules, project context awareness.

My take: Underrated. Catches real issues, not just style nits. The ROI is obvious on any team where PR reviews are a bottleneck.

Codex

OpenAI's autonomous agent. You give it a task, it works in a sandboxed environment, and comes back with a PR.

My take: Promising concept, inconsistent execution. Check back in 6 months.

Lovable

Generates full-stack web apps from descriptions. React + Tailwind + backend in minutes.

My take: Designers love it. Developers should use Claude Code instead — you get better code and more control.

V0

Vercel's UI component generator. Produces Shadcn/ui-compatible React components from descriptions.

My take: Great for one-off component design. I use it occasionally when I want to explore UI ideas quickly before building properly.

Trial Strategy

Don't evaluate everything at once. Four-week plan:

1. Week 1: Set up CodeRabbit on 2–3 active repos. Let it review your PRs for a week. Measure: are reviews faster? Are you catching more?
2. Week 2: Add Copilot. Use it daily. Notice how many suggestions you accept versus dismiss.
3. Week 3: Add Claude Code. Start with a real task — a refactor, a bug fix. Write a CLAUDE.md for your repo. Compare the experience to doing it manually.
4. Week 4: Decide. Keep what made you faster, drop what didn't. Don't pay for overlap.