Which AI coding agent should you use? Pick by budget and use case
The quickest way to pick a coding agent is to start with what you already pay for. ChatGPT, Claude, Google AI and GitHub plans now come with a coding agent, so many builders have one already. Below are the free options and the cheapest paid tiers for each agent, taken from the vendors' own pricing pages, then our take on which one suits which job. For what each tool actually does, read our AI coding agents guide.
Prices are in US dollars per month, before tax. Most plans come with a usage allowance or credits, so "unlimited" rarely means unlimited. Check the linked page before you pay.
Free option · paid plans: ChatGPT Free and Go ($8) for light tasks · Plus $20 · Pro $100, $200 or $500 · Business $20/user (billed annually, 2+ users; $25 monthly)
OpenAI's pricing page says Codex is included in ChatGPT Free, Go, Plus, Pro, Business, Edu and Enterprise plans. Free is pitched for "quick coding tasks" and Go ($8) for "lightweight coding tasks"; both plan cards list a model in the desktop app, subject to rollout. Plus ($20) adds Codex on the web, in the CLI, in the IDE extension and on iOS, plus cloud integrations like automatic code review and Slack. Pro comes in $100, $200 and $500 tiers. You can also run Codex with an API key and pay API rates, which OpenAI calls great for automation like CI, but API-key use has no cloud features such as GitHub code review.
Our take: ChatGPT subscribers, and anyone who wants to script an agent in CI.
Free option · paid plans: Not on Free · Pro $20/month ($17/month billed annually) · Max from $100
Anthropic's plan table lists Claude Code on Pro, Max 5x and Max 20x, but not on the Free plan. Pro costs $20 billed monthly, or $17 a month billed annually ($200 up front). Max starts at $100 a month with 5x or 20x Pro's usage, and Anthropic says Max includes $100 to $200 a month in API credits for building on the Claude Platform. On the Team plan, standard seats are $20 per seat a month billed annually ($25 monthly), and Claude Code is part of the Team plan.
Our take: Long, multi-step changes in an existing codebase, if you already use Claude.
Free option · paid plans: Free: 15 tasks a day, 3 at once · Google AI Pro: 100 a day · Google AI Ultra: 300 a day
Jules's docs list three plans. Free allows 15 tasks in a rolling 24 hours, with 3 running at once. "Jules in Pro" (part of Google AI Pro) allows 100 tasks and 15 at once. "Jules in Ultra" allows 300 tasks and 60 at once. The docs say paid Jules plans currently work only with individual Google accounts ending in @gmail.com.
Our take: Clearing small issues for free while you work on something else.
Free option · paid plans: Free · Pro $10 · Pro+ $39 · Max $100
GitHub's plans page lists Free at $0, Pro at $10, Pro+ at $39 and Max at $100 a month. The Free plan includes agent mode in VS Code, Visual Studio, JetBrains, Eclipse and Xcode, the Copilot CLI and MCP support, with limited chat and agent usage; GitHub says Free users are limited to 2,000 completions a month. Pro adds the cloud agent (assign work to Copilot and it creates a pull request) and pull request code review. Pro+ and Max add delegating tasks to third-party coding agents such as Claude and OpenAI Codex, which GitHub marks as a preview. Agent mode, code review, the cloud agent and the CLI all use GitHub AI Credits.
Our take: The cheapest paid plan here that includes a cloud agent, especially for teams that already work in GitHub.
Free option · paid plans: Free · Plus $8 · Premium $30 · Super $100 · Ultra $200 · Teams from $30/user
Cursor's pricing page now shows individual plans that share one AI usage pool between Cursor and Grok, and the allowance refreshes weekly. Free has limited agent requests. Plus ($8) includes Grok models only. Premium ($30, 4x Plus's usage) adds cloud agents and automations, Bugbot AI code reviews, code hosting, and third-party models at API list price, billed separately by usage. Super ($100) and Ultra ($200) give 15x and 50x Plus's usage. Cursor itself recommends Premium for new AI users and Super for daily agent users, and says Plus is aimed at social creators with limited AI usage.
Our take: Builders who want one editor where agents do most of the typing. If you want non-Grok models, budget for third-party usage on top of the plan.
Free option · paid plans: Free · Pro $20 · Max $200 · Teams $80/month plus $40 per full seat
Cognition's pricing page lists Free (a light agent quota, limited models, unlimited Tab completions and inline edits), Pro at $20 with higher quotas, frontier models and Devin Cloud agents, and Max at $200 with significantly higher quotas. Paid plans include an allowance that refreshes daily and weekly; after that you buy extra usage at API pricing. Windsurf was renamed Devin Desktop on June 2, 2026.
Our take: Ex-Windsurf users, and anyone who wants local and cloud agents in one app for $20.
Free option · paid plans: Free (50 credits) · Pro $20 (1,000 credits) · Pro+ $40 (2,000) · Pro Max $100 (5,000) · Power $200 (10,000)
Kiro's pricing page lists a perpetual Free tier with 50 credits and access to Claude Sonnet 4.5 and open-weight models, with limits. Paid tiers run from Pro at $20 (1,000 credits) to Power at $200 (10,000 credits), and add-on credits cost $0.04 each. Kiro is built by AWS, and its big idea is turning a prompt into a written spec before it writes code.
Our take: Builders who want to plan a feature in writing first, and AWS users.
Free option · paid plans: Free extension and CLI · usage-based model costs, or bring your own key
Cline says its open-source extension is completely free for individual developers. You pay only for AI inference, either through the Cline provider or with your own API keys from providers like Anthropic, OpenAI, Google, OpenRouter, AWS Bedrock and others. Enterprise is a separate paid offering.
Our take: Full control over which model you pay for, inside your editor.
Free option · paid plans: Free, open source (Apache-2.0) · sign in with Google, or use a Gemini API key for usage-based billing
Google's open-source terminal agent. Its README lists three ways to sign in: your Google account, a Gemini API key (usage-based billing, with higher limits when you upgrade), or Vertex AI for enterprise teams. The README doesn't list free-tier quotas, so check them after you sign in.
Our take: Trying a terminal agent for free on Google's models.
Amp's pricing page says the free Hobby tier includes every product feature: you bring your own keys or link your ChatGPT or Claude subscription, self-hosted orbs are free, and Amp-hosted orbs are pay-as-you-go. The $20 Individual tier includes 45,000 minutes of hosted orb time. Amp says it adds no markup to providers' API prices for individuals and non-enterprise workspaces.
Our take: Handing off long tasks to hosted runs and checking back later.
Copilot cloud agent, Codex cloud, Devin Cloud and Jules
You may already have a cloud agent. GitHub's cloud agent comes with Copilot Pro and up, Codex on the web with ChatGPT Plus and up, Devin Cloud with Devin Pro, and Jules has a free tier (all covered above).
Our take: Start with low-risk issues and review every pull request.
Game devs
If you work in Unity, your choice of agent matters less than you might think. Unity offers an official plugin for Claude Code, Codex and Grok, plus a Unity CLI with an MCP server. Unity's FAQ says the CLI, the MCP server and the plugin are all free to use with no concurrency limits. Only Unity's in-editor assistant uses credits, and for Personal users its free trial converts to a $10/month subscription. So pick the agent you already pay for and connect it to the Editor. Unreal and Roblox options are in our AI game dev tools guide.
$0: GitHub Copilot Free in your editor, plus Gemini CLI or Jules Free for terminal and cloud tasks. If you have API credits, use Cline, OpenCode or Aider.
$8 to $10: Cursor Plus ($8, Grok models only) or ChatGPT Go ($8, Codex for lightweight tasks) are the cheapest paid plans here. Copilot Pro ($10) is the cheapest that includes a cloud agent that opens pull requests.
About $20: Pick the ecosystem you already use. ChatGPT Plus gets you Codex, Claude Pro gets you Claude Code, Devin Pro and Kiro Pro are $20, and Amp's Individual tier is from $20.
$100 to $200 (all-day agent use): Claude Max, ChatGPT Pro, Copilot Max, Cursor Super or Ultra, Kiro Pro Max or Power, or Devin Max. Only move up once you're hitting your limits every week.
Teams: Compare per-seat prices on each vendor's team plan. Claude Team, ChatGPT Business, Cursor Teams and Devin Teams list them on the pages linked above.
How to choose (our take)
Already paying for ChatGPT, Claude or Google AI? Use Codex, Claude Code or Jules first. You've already paid for it.
Mostly working in an editor? Copilot if you want a low price and GitHub integration. Cursor if you want one AI-first editor. Devin Desktop or Kiro if you like their cloud-agent or spec-first workflow.
Want open source or a local model? Cline, OpenCode, Aider or Gemini CLI.
Have a backlog of small issues? Add a cloud agent (Copilot cloud agent, Codex cloud, Jules or Devin) and start with low-risk tickets.
Whatever you pick: connect your tools with MCP (see how to add MCP servers), and put your build and test commands in an AGENTS.md file so you can switch agents later without starting over.
FAQ
Which AI coding agent should I use?
Start with the one that comes with a subscription you already pay for (Codex with ChatGPT, Claude Code with Claude Pro, Jules with Google AI). If you have none, GitHub Copilot has a free plan, and Copilot Pro at $10 a month is the cheapest plan here with a cloud agent. Our take: move to a pricier agent only when you hit its limits or need a workflow it lacks.
What's the cheapest AI coding agent?
Several are free. Copilot Free, Jules Free (15 tasks a day), Kiro Free (50 credits), Devin Desktop Free and Cursor Free all have limited free plans. Cline, OpenCode, Aider and Gemini CLI are open source, so you pay only for model usage. The cheapest paid plans in this guide are Cursor Plus and ChatGPT Go at $8 a month, then Copilot Pro at $10.
Is Claude Code free?
No. Anthropic's plan table lists Claude Code on Pro ($20 a month, or $17 a month billed annually) and Max (from $100), not on the Free plan.
Is Codex free?
OpenAI says Codex is included in ChatGPT Free for quick coding tasks. Codex on the web, in the CLI, in the IDE extension and its cloud integrations are listed from Plus ($20). You can also pay API rates with an API key.
Do I need more than one coding agent?
Not to start. Our take: many builders end up with one in the editor and one in the terminal or cloud, but one agent and a good review habit get you further than three agents. See our AI coding agents guide for how each type works.
Why is Cursor's pricing tied to Grok?
Cursor's pricing page shows individual plans that share one usage pool across Cursor and Grok, and the plans also bundle Grok and X features. The $8 Plus plan includes Grok models only; on Premium and above, other providers' models are available at API list price, billed separately. Check the page for the current details.
Every price and feature on this page comes from the vendor's own pricing page or docs (linked as Source) and was checked on October 21, 2026. Prices change often, so check the link before you pay. There are no affiliate links on this page.
How we write these: each tool description is based on the tool's own site or docs (linked as Source) and checked on the date above. "Our take" lines are our opinion. Something out of date? Email hello@hotfixdaily.com. No tool paid to be listed.
Get the next issue in your inbox
Five things worth your attention in AI, agents, vibe coding and game dev, every weekday. Plus the free AI Builder Starter Stack PDF when you join.
