Anthropic Quietly Removes Claude Code From the $20 Pro Plan — Here’s What It Means

If you logged into Claude this week and saw Claude Code missing from the Pro plan’s feature list, you weren’t imagining it. Anthropic’s pricing page on April 21, 2026 quietly started showing an X next to Claude Code under the $20/month Pro tier, with checkmarks only under Max 5x ($100/month) and Max 20x ($200/month). The catch: not every Pro user is seeing the change, and according to Anthropic, most of them shouldn’t be seeing it at all.

This is a split test, not a universal rollout — at least for now. Here’s exactly what’s happening, who’s affected, and what to do if Claude Code suddenly disappears on your account.

What Anthropic Actually Said

Anthropic’s head of growth Amol Avasare addressed the firestorm directly on X, saying the company is “running a small test on ~2% of new prosumer signups” and that “existing Pro and Max subscribers aren’t affected.” That’s the official position as of April 22, 2026.

Translated: if you signed up for Pro before this week, your Claude Code access should still work. If you’re brand new to Pro and you got pulled into the unlucky 2%, you’re seeing the Max-only version of the pricing page and you don’t get Claude Code at the $20 tier.

The Register independently confirmed the inconsistency — when their reporter accessed Claude Code from the CLI on an existing Pro account, the terminal still printed “Claude Pro” and the tool worked normally. The Claude Code product page on Anthropic’s site also still listed Pro as having access at the time the story broke. So the split test is real, but Anthropic’s own site hasn’t been updated consistently to reflect that it’s a test rather than a permanent change.

Why This Got Misread as a Full Removal

A few things made this look like a sitewide policy change rather than a 2% experiment:

The main pricing page got updated with the X mark. Anyone landing on claude.com to compare plans saw Claude Code gated to Max — no banner, no asterisk, no “this is a test” disclosure.

The support documentation was also edited. The page that previously read “Using Claude Code with your Pro or Max plan” was rewritten to “Using Claude Code with your Max plan.” That’s the kind of edit you make when a decision is final, not when you’re A/B testing.

These two changes hitting at the same time, with no announcement and no email to subscribers, made it look like Anthropic had quietly removed the feature and was hoping nobody would notice. Hours later, Avasare’s tweet reframed it as an experiment — but the public-facing site still doesn’t say that.

How to Tell If You’re In the Test

Check three places:

Your pricing page view. Log in, go to claude.com/upgrade, and look at the Pro column. If Claude Code shows an X, you’re either in the test or viewing the test variant. If it shows a checkmark, you’re on the legacy view.

Your CLI. Open a terminal and run claude (assuming you have Claude Code installed). If it authenticates and the header shows “Claude Pro,” you still have working access regardless of what the pricing page says.

Your account settings. Inside claude.ai, the plan details panel should reflect what your subscription actually includes — this is more reliable than the marketing page.

If the CLI works for you, you have access right now. That’s the only test that matters.

Why Anthropic Is Running This Experiment

Avasare was unusually candid about the underlying problem in his X thread. His exact framing: “When we launched Max a year ago, it didn’t include Claude Code, Cowork didn’t exist, and agents that run for hours weren’t a thing. Max was designed for heavy chat usage, that’s it.” Since then, Claude Code launched, got bundled into Max, exploded in popularity after Opus 4 and 4.7, and long-running agents became a normal workflow instead of a novelty.

The economics shifted. A $20 Pro user running Claude Code sessions all day on Opus 4.7 can easily consume more compute than $20 covers — by a factor of 10 or more in heavy cases. Anthropic introduced weekly caps and tighter peak-hour limits earlier this year to manage this, but the gap between subscription revenue and inference cost on the heaviest users keeps widening.

This isn’t unique to Anthropic. Last week, internal documents showed Microsoft is shifting GitHub Copilot toward token-based billing for similar reasons. Every AI coding tool that priced its subscription before the agentic era is now reckoning with the same math.

What Your Options Are If You Lose Access

If Claude Code disappears from your Pro plan — whether through this test or a future expansion of it — here’s the practical breakdown:

Stay on Pro at $20/month. You keep Claude on web, iOS, Android, and desktop. You keep Projects, Artifacts, web search, file uploads, and access to Sonnet 4.6 and Opus 4.7 in the chat interface. You lose the terminal coding agent.

Upgrade to Max 5x at $100/month. Restores Claude Code plus 5x the usage capacity of Pro. This is the new floor for terminal access if the test rolls out broadly.

Upgrade to Max 20x at $200/month. Same access, 20x Pro capacity. Worth it only if you’re running Claude Code heavily across full workdays.

Switch to API billing. Claude Code works with API credits. Sonnet 4.6 runs $3 per million input tokens and $15 per million output tokens; Opus 4.7 is more expensive. Light-to-moderate users may spend less than $100/month on actual usage. The Batch API offers a 50% discount for non-time-sensitive work. Check your token consumption history before committing either way.

Look at competitors. GitHub Copilot, Cursor, Cody by Sourcegraph, Continue.dev, and OpenAI’s Codex (which is currently still bundled in the $20 ChatGPT Plus tier — OpenAI’s Tibo Sottiaux made a point of this on X in response to the Anthropic news) are all viable alternatives at lower price points. None of them match Claude Code’s specific agentic terminal workflow, but they get the job done for most coding tasks.

Run a local model. Llama-based and Qwen-based coding models running on a decent GPU have closed enough of the gap that for routine code generation on familiar codebases, they’re a serious option. This isn’t a fit for everyone, but the cost ceiling is whatever your hardware costs to run.

Should Existing Pro Users Worry?

Probably yes, eventually. Here’s the honest read:

A 2% test is how a permanent change usually starts. Companies run small experiments, measure conversion to higher tiers versus churn, and if the math works, they roll the change out broadly. The fact that Anthropic also updated the global support documentation — not just the test variant — suggests the internal decision direction is already set, even if the formal rollout hasn’t happened.

Avasare himself said in the same thread that Anthropic is “looking at different options to keep delivering a great experience for users” and “we don’t know exactly what those look like yet — that’s what we’re testing and getting feedback on right now.” That’s not the language of a company committed to keeping Claude Code on Pro indefinitely. It’s the language of a company figuring out how to move it without losing too many customers.

If you depend on Claude Code at the $20 tier, the responsible move is to start tracking your actual API token consumption now so you know whether Max 5x, API pay-as-you-go, or a competitor makes more sense for your workload when the change does roll out.

The Communication Problem

The thing developers are most upset about isn’t the price — it’s the silence. There was no email, no in-product notification, no blog post, no changelog entry. The only signal was the pricing page edit and the documentation rewrite. Avasare’s clarification came after journalists like Ed Zitron flagged the change, not before.

For a $20/month consumer product, a small split test probably doesn’t warrant a press release. But updating the global support documentation to remove the word “Pro” — while telling the public that Pro users are unaffected — is the kind of contradiction that erodes trust faster than the actual policy change does. If the test results in a real removal down the road, customers who already feel they were misled by the documentation update will be harder to retain.

Bottom Line

Right now, if you’re an existing Pro subscriber and Claude Code still works for you in the CLI, nothing has changed. You’re not in the test group, and Anthropic has explicitly said you won’t be affected.

If you’re a new Pro signup and Claude Code is missing from your plan, you’re one of the 2% in the experiment. Your options are to upgrade to Max, switch to API billing, or move to a competitor.

If you’re shopping the Pro plan specifically because of Claude Code, hold off on subscribing until Anthropic publishes a clear policy. Either you’ll get the legacy view with access included, or you won’t — and the documentation right now isn’t reliable enough to bet $20 on without checking your account first.

This is a developing situation. Anthropic hasn’t issued a formal statement beyond Avasare’s X posts, and the inconsistency between the pricing page, the product page, the support docs, and the CLI suggests the company is still working out internally how to handle this. Expect more clarity — or more confusion — over the coming days.