First-hand: this site is built with these tools

AI Coding Tools in 2026: What to Actually Use

Three families, one decision: terminal agents you delegate whole tasks to, AI IDEs you drive interactively, and extensions that upgrade the editor you already use. We use these daily — the verdicts below are working opinions, not spec-sheet rewrites.

ToolTypePricing (mid-2026)ModelsMCPOpen source
Claude Code
Anthropic
Terminal agentFrom $20/mo (Claude Pro); Max plans $100–200/mo; or pay-as-you-go APIClaude (Sonnet/Opus/Fable tiers)
Cursor
Anysphere
AI IDEFree tier; Pro ~$20/mo; Ultra ~$200/moMulti-model: Claude, GPT, Gemini + own fast models
Windsurf
Windsurf (formerly Codeium)
AI IDEFree tier; Pro ~$15/moMulti-model incl. own SWE models
GitHub Copilot
GitHub / Microsoft
Editor extensionFree tier; Pro ~$10/mo; Pro+ ~$39/moMulti-model: GPT, Claude, Gemini (selectable)
Codex CLI
OpenAI
Terminal agentIncluded with ChatGPT Plus/Pro plans; or API pay-as-you-goGPT / o-series (OpenAI only)
Cline
Open source community
Open-source agentFree (BYO API key — you pay token costs directly)Any: Claude, GPT, Gemini, DeepSeek, local via Ollama/OpenRouter
Aider
Open source (Paul Gauthier)
Open-source agentFree (BYO API key)Any via API: Claude, GPT, Gemini, DeepSeek, local models

Head-to-head comparisons

Guides

Subscription or API?

Rule of thumb from our own bills: casual use (<1h/day) — a $10–20 subscription wins. Daily agentic work — flat subscriptions (Claude Pro/Max, ChatGPT plans) beat raw API billing by a wide margin. Cost-optimizing with open-source agents (Cline, Aider)? Pair them with a value-leader model from our Value Leaderboard and estimate real usage on the token cost calculator.

Disclosure: no affiliate links anywhere in the coding hub. Verdicts are our own working opinions from daily use.