The best AI code tools: coding agents, editors and assistants

AI coding tools now come in three shapes: an agent you run in the terminal, an editor with AI built in, and an assistant that plugs into the editor you already use. Most of them blur these lines, and all of them do more than an AI code generator that finishes one line at a time. This page compares seven tools on price, free tier and where they run, using each vendor's pricing page as of October 2026, so you can pick one to try this week.

How the picks were chosen

We kept AI programming tools that can read a codebase, edit files and run commands or tests. Each pick publishes its prices, runs in a mainstream editor or the terminal, and is sold to individual developers as well as companies. Criteria come from public facts: pricing, features listed by the vendor, and supported platforms. We did not run our own benchmark.

Two practical notes. Every tool here meters usage with credits, quotas or weekly limits, so the plan price is the minimum you will spend. And the model behind the tool matters as much as the tool: several of these let you pick models from more than one lab.

The seven tools at a glance

1. Claude Code

Best for: multi-file changes and long tasks you steer from the terminal · Price: included with Claude Pro at $20/month, or $17/month billed annually (as of Oct 2026) · Free plan: no

Claude Code is Anthropic's coding agent. Its docs describe it as a tool that reads your codebase, edits files, runs commands and works with your development tools. It runs in the terminal on macOS, Linux and Windows, in VS Code and JetBrains, as a desktop app, on the web and in Slack. Claude's pricing page lists Claude Code from the Pro plan up; Max starts at $100 a month for 5x or 20x Pro usage.

Limits: there is no free tier for Claude Code, and heavy daily use pushes you toward Max. For setup and commands, keep our Claude Code cheat sheet open, and use the Claude Code prompts for common tasks.

2. Cursor

Best for: an AI-first editor with agents and code review · Price: Individual (Pro) $20/month; Teams $40/user/month (as of Oct 2026) · Free plan: yes

Cursor is an editor built around AI agents. The free Hobby plan needs no credit card and includes limited Agent requests and access to Composer. The Individual plan starts at $20 a month for Pro, with Pro+ and Ultra tiers above it, and adds extended Agent limits, frontier models, MCPs, skills and hooks, and cloud agents. Cursor recommends Pro+ for daily agent users and Ultra for power users. Its Bugbot code review runs on usage-based billing.

Limits: every plan includes a set amount of model usage, and work beyond it is billed on demand. Privacy mode, which Cursor says guarantees code is not used for training, has to be switched on. For a direct comparison, read Claude Code vs Cursor.

3. GitHub Copilot

Best for: developers who live in GitHub and want the cheapest paid plan · Price: Pro $10/month; Pro+ $39/month; Max $100/month (as of Oct 2026) · Free plan: yes

GitHub Copilot works in VS Code, in the Copilot CLI, in a desktop app and in agents on GitHub itself. The Free plan gives limited chat and agent usage and supports the CLI and agent mode. Paid plans are measured in GitHub AI Credits, which chat, agent mode, code review, the cloud agent and the CLI all consume. Pro+ and Max add larger credit allowances.

Limits: usage is credit-based, so a busy week of agent work can use up a plan faster than chat alone.

4. OpenAI Codex

Best for: trying an agent for free, and ChatGPT subscribers · Price: included with ChatGPT; Go $8/month, Plus $20/month (as of Oct 2026) · Free plan: yes

Codex is OpenAI's coding agent, included in every ChatGPT plan. The Free plan gives GPT-6 Luna in the desktop app, subject to rollout. Plus at $20 a month adds Codex on the web, in the CLI, in the IDE extension and on iOS, plus automatic code review and a Slack integration, with GPT-6.1 Sol. Pro plans cost $100, $200 or $500 a month. You can also run Codex with an API key and pay API rates, which suits CI. Business costs $20 per user per month billed annually, with no training on business data by default.

Limits: subscription usage is shared between Codex and ChatGPT Work. OpenAI says GPT-5.5 retires from Codex on October 14, 2026. Compare it head to head in Claude Code vs Codex.

5. Devin Desktop (formerly Windsurf)

Best for: an editor with unlimited tab completion and access to cloud agents · Price: Pro $20/month; Max $200/month (as of Oct 2026) · Free plan: yes

Cognition introduced Devin Desktop on June 2, 2026 as the next generation of the Windsurf editor, and windsurf.com now redirects to Devin's site. The Free plan has a light agent quota, limited models, and unlimited inline edits and Tab completions. Pro adds bigger quotas, models from OpenAI, Claude, Gemini and SpaceXAI plus leading open-source models, and access to Devin Cloud agents. Teams cost $80 a month for the team plus $40 per full developer seat.

Limits: quotas refresh daily and weekly; past them you buy extra usage at API pricing. Cognition's own SWE-2 model is free in Devin Desktop and the CLI only through October 16, 2026.

6. Google Antigravity

Best for: Google AI subscribers and agent-first workflows · Price: $0 for individuals; higher limits with Google AI Pro at $19.99/month (as of Oct 2026) · Free plan: yes

Antigravity is Google's agent platform for developers, with an app, a CLI, IDE extensions and an SDK. The $0 individual plan includes a variety of agent models, unlimited Tab completions and command requests, and basic weekly rate limits. Google AI Pro and Ultra raise those limits and add a flexible AI credit pool. Companies can buy it through Google Cloud, with Gemini Enterprise seats from $30 per user per month.

Limits: the free plan's weekly rate limits are described only as basic, so expect to hit them on long sessions.

7. Mistral Vibe

Best for: a terminal agent from a vendor with open-weight models · Price: Pro $14.99/month before tax; verified students $5.99 (as of Oct 2026) · Free plan: yes

Mistral sells its coding agent as Vibe for code. The Vibe CLI is a terminal agent powered by Mistral's Devstral model, and the IDE plugin for VS Code and JetBrains adds tab completion and chat. The Free plan includes limited coding sessions. Pro adds all-day coding in the CLI, IDE or web. Mistral lists self-hosted, private-cloud or on-premises deployment among its team and enterprise features.

Limits: Pro runs under fair-use limits, and the Free plan is meant for testing.

Which one to pick

• You want an agent that works across a whole repo from the terminal: Claude Code or Codex.
• You want AI built into the editor: Cursor or Devin Desktop.
• You want the lowest cost: GitHub Copilot Free, Antigravity's $0 plan, or Codex on ChatGPT Free.
• You need code review in pull requests: Copilot, Cursor's Bugbot or Codex on Plus all offer it.
• You build on a model API yourself: see OpenAI API alternatives.

Whatever you pick, the agent works better when it plans before it edits. This prompt works in any of them:

Read [file or folder]. Explain what it does in five bullet points.
Then propose a plan to [change you want]: list the files you will touch, the tests you will run, and any risk.
Do not edit anything until I reply "go".

For prompts that are not tied to one tool, browse our coding prompts.

FAQ

What is the best AI for coding and programming?

It depends on how you work. Agents such as Claude Code and Codex suit multi-file changes; editors such as Cursor and Devin Desktop suit people who want AI in every keystroke; GitHub Copilot suits teams already on GitHub. Try two free tiers on the same task.

Which AI programming tools are free?

Cursor Hobby, GitHub Copilot Free, Codex on ChatGPT Free, Devin Desktop Free, Antigravity's $0 plan and Mistral's Free plan. All have tight usage limits. Claude Code starts with the paid Pro plan.

Will my code be used to train models?

It depends on the tool and plan. Cursor says code is not used for training when privacy mode is on. OpenAI says Codex on Business does not train on business data by default. Read the data page for the plan you buy.

What happened to Windsurf?

Cognition introduced Devin Desktop on June 2, 2026, describing it as the next generation of Windsurf. Windsurf.com now redirects to Devin's website.