Claude Code Alternative: 6 Tools, 4 With a $0 Plan

In brief

Codex and Google's Antigravity CLI are the closest alternatives to Claude Code, as of November 2026. Artificial Analysis scores Antigravity CLI 55 and Codex 53 against Claude Code's 58, at $7.13 and $7.63 a task against $15.47. Both have a $0 plan, and Claude Free doesn't include Claude Code.

What are the best alternatives to Claude Code?

Six tools replace Claude Code in a terminal today: Codex, Antigravity CLI, GitHub Copilot CLI, OpenCode, Cline and Aider. We picked them because each one runs from a shell and each one has a page you can check. We read each vendor's own plans page or GitHub repo on 14 November 2026. Claude Code itself costs $20 a month on Pro, or $17 a month billed as $200 up front, and Claude Free doesn't include it, per Claude's pricing page.

ToolWhat it costs, vendor's pageModels it runsLicense and GitHub starsPick it if
Codex (OpenAI)Free $0, Go $8, Plus $20, Pro from $100GPT-6.1 Sol and GPT-6 Luna on PlusApache-2.0, 135.3k starsYou already pay for ChatGPT
Antigravity CLI (Google)Individuals $0, Google AI Pro $19.99Gemini 4 models, including HydrogenNot stated on its plans pageYou want a free CLI with a weekly quota
GitHub Copilot CLIFree $0, Pro $10, Pro+ $39, Max $100Model picker with /model, plus your own providerNot stated on its docs pageYou already pay GitHub $10
OpenCodeFree models with no account, Go $10200+ providers through Models.devMIT, 227K starsYou want any model in one terminal
ClineNot stated on its repo pageAny provider you connectApache-2.0, 71.6k starsYou want the same agent in VS Code and the CLI
AiderNot stated on its siteAlmost any LLM, including local modelsApache-2.0, 50.1k starsYou want every edit committed to git
A terracotta Claude starburst on a tall black plinth in the middle, with a brushed silver OpenAI knot on the left and a rainbow glass Gemini star on the right, each on a low plinth in a dark studio with a thin line of white light along the floor

GitHub describes Codex as a "Lightweight coding agent that runs in your terminal," and describes Aider as "AI pair programming in your terminal." OpenCode calls itself "The open source coding agent for any model." Cline calls itself an "Autonomous coding agent as an SDK, IDE extension, or CLI assistant." Each of those lines comes from the tool's own page, so it tells you what the project says about itself and not how it performs.

The price column is a floor, not a bill, and it hides some detail. Codex prices come from ChatGPT pricing. Codex lists its CLI under Plus, and its Free and Go rows name the desktop app. Cline and Aider are open source, but both need a model behind them, and the model is where the bill comes from. If your objection is the terminal itself, Cursor vs Claude Code covers the IDE route.

How close do the alternatives get to Claude Code?

Antigravity CLI gets within 3 points of Claude Code and Codex within 5, as of 14 November 2026. Artificial Analysis runs each agent on the same tasks and averages three tests into its Coding Agent Index v2.1. Claude Code with Claude Fable 5.2 at Xhigh effort scores 58 and costs $15.47 a task. Here's every run that matters for a switch:

Claude Code on Medium ties Codex's best score for $2.89
0153045601251015API cost per task, US dollars, Artificial Analysis, read 14 Nov 2026Coding Agent Index v2.1OpenCode, GLM-5.3Codex, GPT-6.1 SolClaude Code, Fable 5.2 MediumCodex, GPT-6 AstraAntigravity CLI, Gemini 4 CarbonClaude Code, Fable 5.2 HighClaude Code, Fable 5.2 Xhigh

The chart changes the story most guides tell. Claude Code at High effort scores the same 58 for $7.51 a task, 12 cents less than Codex's best run at $7.63. So the headline gap of $15.47 against $7.63 is a choice of effort setting, not a fixed price. Codex does win on speed: 26.0 minutes a task for GPT-6 Astra against 53.3 for Claude Code at Xhigh. And Codex with GPT-6.1 Sol scores 52 for $1.51, a tenth of the top Claude Code run.

Agent, best runCoding Agent IndexFrontierCodeTerminal-BenchSWE-Atlas QnACost per task
Claude Code, Fable 5.2 Xhigh58496462$15.47
Antigravity CLI, Gemini 4 Carbon55386564$7.13
Codex, GPT-6 Astra53435859$7.63
OpenCode, GLM-5.340214059$5.41

If you read only one column, read FrontierCode, because that is the test where the gap is widest. Claude Code wins it by a wide margin: 49 on FrontierCode, against 43 for Codex, 38 for Antigravity CLI and 21 for OpenCode. Antigravity CLI edges it on the other two, 65 to 64 on Terminal-Bench and 64 to 62 on SWE-Atlas QnA. Two caveats matter. Artificial Analysis prices every run at API rates, so it says nothing about a flat plan. And it ran OpenCode with one model, GLM-5.3, so that 40 scores the model as much as the tool. We opened three of its comparison pages (Codex, Antigravity CLI, OpenCode). None we opened scores Copilot CLI, Cline or Aider, so we give no score for them. Claude Code vs Codex goes deeper on that pair.

Is there a free alternative to Claude Code?

Yes, four of the six have a $0 plan, and two of them put the CLI on it: Antigravity CLI and GitHub Copilot CLI. Google's plans page says free users get "A meaningful quota, refreshed weekly," and lists "Access to all product features, such as scheduled tasks and the CLI" on every plan. Google prints no quota number on that page, so we can't tell you how much work a free week holds. Our Antigravity free page tracks what it does say.

GitHub's Copilot plans page gives Free users 2,000 completions a month and "Limited chat and agent usage," and says the Free plan supports the CLI and agent mode. Pro costs $10 and adds $15 in monthly AI credits, where one credit equals one cent. OpenCode offers free models with no account, and its Go plan costs $10 a month with a dollar allowance per model, across 32 models. Codex Free and Go list GPT-6 Luna in the desktop app. Artificial Analysis scores GPT-6 Luna at 26 in Codex, so check that it fits the work before you count it as a replacement.

Google also changed where its older tool lives. The Gemini CLI repo says: "Free individual access and Google AI Pro / Ultra (Google One) moved to Antigravity CLI on June 18, 2026." So the Gemini CLI you may see in older guides no longer carries free individual access. It's Apache 2.0 with 108.8k stars.

What if the cap is the problem, not the tool?

If you hit a usage cap and just need coding help, Whizi's free plan covers questions you paste in. It 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. Whizi is a chat app, not a terminal agent, so it won't edit your repo. Pro carries Claude, GPT and Gemini in one place, and its first week is $1.99.

That fits a narrow job, and it's one that people do all day. You paste an error, a function or a diff, and ask what's wrong. It doesn't replace the agent loop of reading files, running tests and committing, which is the part of the job that the six tools above do for you.

Can you keep Claude Code and swap the model?

Yes. Claude Code can run on open models through Ollama, and that is the cheapest way to keep its interface. Ollama's guide gives one command, ollama launch claude, or you can point ANTHROPIC_BASE_URL at a local server. It names a hard limit: "For larger repositories, set the context length to 64k or higher." It also says hosted WebSearch and advanced tool controls aren't fully supported.

The score drops with the model. Artificial Analysis lists Claude Code on Qwen3.8 Max at 33 and on Step 5 Preview at 36, against 58 on Fable 5.2 Xhigh. Haiku 5.5 on Max effort scores 30. That points at the model, not the harness. It's also why OpenCode's 40 on GLM-5.3 says little about OpenCode.

When is Claude Code still the better pick?

Claude Code is still the better pick when the job is hard. It leads FrontierCode by 6 points over Codex, the largest gap on any test, and Pro at $20 includes it under usage limits instead of a per-task meter. Claude Code pricing lists every plan. The cheapest way to use Claude Code is Pro billed yearly.

You arePick
Already on ChatGPT PlusCodex, at no extra cost
Looking for a free CLIAntigravity CLI, with a weekly quota
On GitHub alreadyCopilot CLI, from $0 or $10
Wanting one terminal for any modelOpenCode
Doing the hardest refactorsClaude Code, which led FrontierCode
Common questions

Frequently asked questions

What is the best alternative to Claude Code?

Codex is the best-known alternative and Antigravity CLI is the closest on score. Artificial Analysis gives Claude Code 58, Antigravity CLI 55 and Codex 53, as of 14 November 2026. OpenAI lists Codex access at $8 on Go and $20 on Plus. Codex pricing has every plan.

Is there an open source alternative to Claude Code?

Yes. OpenCode is MIT licensed, and Codex, Cline and Aider are Apache-2.0, per each project's own page. OpenCode has 227K GitHub stars. All four need a model behind them, which can be a local one, so the software is free and the model may not be.

What is Google's alternative to Claude Code?

Antigravity CLI. Google's plans page lists it on every tier, from the free Individuals plan to Google AI Ultra, and Google AI Pro costs $19.99 a month in the US. Gemini CLI's repo says free individual access moved to Antigravity CLI on June 18, 2026. Antigravity pricing lists every plan.

Is GitHub Copilot a Claude Code alternative?

Partly. Copilot CLI runs in the terminal and its Free plan marks CLI and agent mode as supported, with "Limited chat and agent usage." Pro costs $10 and includes $15 in AI credits. GitHub's docs list no per-plan CLI caps. Is GitHub Copilot free covers the Free plan.