Codex alternatives: Copilot Pro runs a cloud agent for $10

In brief

Claude Code is the closest Codex alternative, and GitHub Copilot Pro is the cheapest one with both a cloud agent and a CLI, at $10 a month against $20 for ChatGPT Plus, as of November 2026. Antigravity and OpenCode cost $0 to start. Cursor's cloud agents start at $30.

Which tools replace Codex, and what does each cost?

Six tools replace OpenAI Codex, and five of them start at $20 a month or less. We read each price on the vendor's own pricing page on 14 November 2026. US prices, before tax. Codex itself comes with ChatGPT Plus at $20 a month (ChatGPT pricing) for the web app, CLI, IDE extension and iOS, and with GPT-6 Luna in the desktop app at $0.

A brushed silver OpenAI knot on a tall black plinth, with a terracotta Claude starburst, a rainbow glass Gemini star and a frosted glass terminal cube on lower plinths around it in a charcoal studio.
AlternativeCheapest plan that carries itCloud agentCLIModels it runs
GitHub CopilotPro, $10 a monthYes, Pro and upYes, on every planA model picker, billed in AI credits
Claude CodePro, $20 a month, or $200 billed yearlyYes, cloud sessions on Pro and upYesClaude models
DevinPro, $20 a monthYes, Devin CloudNot stated on its pricing pageFrontier models, extra usage at API pricing
CursorPremium, $30 a monthYes, Premium and upLinked, plans not statedOnly Grok models on Plus, third-party models at API list price
AntigravityIndividuals, $0Not listedYes, on every planGemini, plus Claude models on some plans
OpenCodeFree models, $0Not listedYes200+ providers, including local models
Copilot Pro carries a cloud agent for half of Codex Plus
Codex on ChatGPT Plus$20 a month$20GitHub Copilot Pro$10 a month$10Claude Pro$20 a month$20Devin Pro$20 a month$20Cursor Premium$30 a month$30each square is one US dollar of themonthly plan price, read 14 November 2026

The chart leaves out Antigravity and OpenCode because they have nothing to count. Both start at $0. Every row still has a catch, and the next sections go through them one at a time.

What is the cheapest Codex alternative with a cloud agent?

GitHub Copilot Pro is the cheapest at $10 a month, and it keeps the cloud agent and the CLI. GitHub's plans page lists the coding agent on Pro and Pro+ and leaves it off Free. Its billing docs say "All plans include Copilot CLI and the Copilot app." You hand the agent a task and it opens a pull request.

Usage is metered in credits. GitHub's individual plans docs give Pro 1,500 AI credits a month, made of 1,000 base and 500 flex, and the plans page says "1 AI credit = $0.01 USD." That makes $15 of model usage inside a $10 plan. When the credits run out you can wait, set a dollar budget for more, or pick a cheaper model. Pro+ is $39 and carries 7,000 credits.

Here's where Codex wins. OpenAI sizes Plus in messages, 15 to 160 GPT-6.1 Sol messages per 5 hours, and Copilot sizes Pro in dollars, so the two can't be set side by side without a guess. We won't make one. The Plus card reads "Power a few focused coding sessions each week." A cheaper bill is a reason to try Copilot for a month, not proof that it does more work per dollar.

GitHub also lists Claude and Codex as third-party agents inside Copilot. The plans page contradicts itself on the plan: the Pro list says "Access to 3rd party agents (Claude Code and Codex)," while the comparison table marks delegating to them as Preview on Pro+ only. Check which one your account shows before you count on it.

Is Claude Code a good Codex alternative?

Claude Code is the closest match to Codex, because it comes as a terminal agent and as a cloud agent for the same $20. Anthropic's pricing page marks Claude Code as not included on Free and included on Pro, which is $20 a month, or $17 a month billed as $200 up front. Max starts at $100 a month.

The cloud half works like Codex's cloud tasks. Anthropic's cloud sessions docs say sessions are "available on Pro, Max, Ultra, and Team plans," and that a session "counts toward your plan's usage limits alongside the rest of your Claude and Claude Code usage, and there's no separate charge for the cloud VM." You can start one from a browser, the phone app or claude --cloud.

The catch is the shared allowance. Claude chat and Claude Code draw from the same limits, so a long chat in the morning leaves less for the terminal. Claude Code only runs Claude models, and Codex only runs OpenAI's. If you want to know which agent does better work on your repo, Claude Code vs Codex sets the two side by side, and Claude Code pricing has every plan.

What is the best free Codex alternative?

Antigravity and OpenCode are the two free terminal agents, and Codex's own free tier is the smallest of the three. Free Codex gives you GPT-6 Luna in the ChatGPT desktop app, "subject to rollout," and OpenAI prints no usage range for it. Is Codex free has the whole list.

Google's Antigravity plans page says accounts not on Google AI Pro or Ultra get "A meaningful quota, refreshed weekly," and that every plan gets "Access to all product features, such as scheduled tasks and the CLI." Its models page lists Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS-120b for free accounts and adds "Will be removed on November 18, 2026" next to them. What you get on a free plan is changing this month. Google doesn't publish the quota as a number. Our Antigravity pricing page tracks it.

Google's Gemini CLI is the old free route, and it has moved. Its GitHub page says "Free individual access and Google AI Pro / Ultra (Google One) moved to Antigravity CLI on June 18, 2026." The repository still has 108.8k stars and an Apache 2.0 license, so the code is open, but free sign-in lives in Antigravity now.

OpenCode needs no account for its free models. Its site says "OpenCode comes with a set of free models. No account needed." It also sells OpenCode Go, "curated models from $10/month." The site doesn't name the free models or their limits on its front page, so check the list before you rely on one.

Free plans run out, and that's when the question changes from which agent to where else you can ask for help. Whizi, which publishes this page, has no coding agent and won't edit your repo. Its free plan gives 10 messages a day on Whizi AI, Auto and Claude Haiku 5.5, 2 pictures a month and 5 photo uploads a day, with no card. That's enough to ask a chat model about an error you pasted in. Pro carries Claude, GPT and Gemini, and its first week is $1.99.

Is there an open source alternative to Codex?

Yes, OpenCode is the biggest one, and it has more GitHub stars than Codex itself. Its repository shows 226.9k stars and an MIT license on 14 November 2026. OpenAI's openai/codex repository shows 135.3k stars and Apache 2.0. So Codex's CLI is open source too. What OpenAI doesn't open is the model behind it.

OpenCode's site says it supports "200+ LLM providers through Models.dev, including local models," and that you can "Log in with ChatGPT Plus or Pro or GitHub Copilot." That makes it a front end for plans you may already pay for. A Plus subscriber can try a second agent without a second bill. The page doesn't list a Claude login, so Claude Code is the only way in for a Claude plan.

Running a local model with OpenCode costs nothing to the vendor, but it costs you speed. A model small enough for a laptop is slower and weaker than GPT-6.1 Sol on a long job across many files. For a local run of Codex itself, the --oss route in Codex pricing works the same way.

What about Cursor and Devin for cloud agents?

Cursor starts its cloud agents at $30 a month, and Devin starts at $20. Cursor's pricing page lists "Cloud agents & automations" on Premium and higher, and its Plus plan at $8 holds only Grok models. Third-party models bill separately at API list price. Codex vs Cursor works out a month of tasks on both.

Devin's pricing page lists Free at $0, Pro at $20, Max at $200 and Teams from $80. Pro includes Devin Cloud and the option to "Purchase extra usage at API pricing." It publishes no quota size, and its Free plan carries only a "Light quota." Pick Devin if you want a cloud agent more than an editor.

Which Codex alternative should you pick?

Pick by what you want to stop paying for, not by feature count. Here is the rule we'd use, from the prices above as of 14 November 2026:

If you wantPickMonthly cost
A cloud agent and a CLI for the least moneyGitHub Copilot Pro$10
The nearest thing to Codex on a different lab's modelsClaude Code on Claude Pro$20
A cloud agent that isn't tied to a chat planDevin Pro$20
An editor first, agents secondCursor Premium$30
A free terminal agent from GoogleAntigravity$0
A free open source agent and your choice of modelOpenCode$0
OpenAI's newest models in one planStay on Codex, ChatGPT Plus$20

Codex still wins in three places. It runs OpenAI's newest models, GPT-6.1 Sol and GPT-6 Astra among them. It puts the web app, the CLI, the IDE extension and the phone on one $20 plan. Every alternative here is cheaper only if you give up one of those.

Workflow checklist
  • Try GitHub Copilot Pro at $10 for a month if you want a cloud agent and a CLI for less than Plus
  • Install Claude Code on Claude Pro if you want the nearest match from another lab
  • Run OpenCode on its free models before you pay for anything
  • Keep Codex on Plus if GPT-6.1 Sol is the model you need
Common questions

Frequently asked questions

Is there a free alternative to Codex?

Yes, two. Antigravity gives accounts without Google AI Pro "A meaningful quota, refreshed weekly" with the CLI on every plan, and OpenCode ships free models with no account. Neither publishes a number. Free Codex is GPT-6 Luna in the desktop app. Is Codex free lists every free route.

Is there a Codex alternative from Google?

Yes, Antigravity. Google moved free individual access from Gemini CLI to Antigravity CLI on 18 June 2026, according to the Gemini CLI repository. Google AI Pro adds "A high, generous quota, refreshed every five hours until the weekly limit is reached," and Antigravity's page lists no price for it. Antigravity pricing compares the plans.

Can I use Codex inside another tool?

Yes, in two ways. OpenAI's Codex extension installs in Cursor and other VS Code editors and bills your ChatGPT plan. GitHub lists Codex as a third-party agent in Copilot (Pro+ at $39 a month is the plan its comparison table names), and OpenCode says you can log in with ChatGPT Plus or Pro.

Is Claude Code better than Codex?

It depends on the repo, and neither vendor publishes a head-to-head. Both cost $20 on their cheapest plan with a terminal agent and a cloud agent. Claude Code vs Codex compares the plans, the limits and the scores.