All offers / AI tools / Is Claude free?
Yes, Claude is free, and the free plan is a product rather than a demo.
Anthropic puts a price on the Free plan and no figure at all on the allowance. The price is the $0 above; nothing on that pricing page, in the consumer terms effective October 8, 2025, or in the four help-centre articles read alongside them says how much free actually is. All of them were fetched live on July 27, 2026, and every one answered HTTP 200.
What the free plan actually gets you
Anthropic's plan comparison marks the Free column as included for chat on web, iOS, Android and desktop, web search, desktop extensions, voice mode, incognito chats, user preferences, artifacts, creating and editing files with code execution, memory, skills and connectors. The Free plan card adds "Connect Slack and Google Workspace services", "Integrate any context or tool through connectors with remote MCP" and "Extended thinking for complex work".
Memory across conversations is on the free plan, and so are connectors, which is how Claude reaches your other tools.
And precisely where it stops
Anthropic marks the exclusions with a cross, not a footnote, which is unusually legible of them. On the Free column, these are crossed out:
| Claude Code, Claude Cowork, Claude Design, Claude Science | Not on Free |
|---|---|
| Projects | Not on Free |
| Research | Not on Free |
| Claude for Chrome, Microsoft 365, Outlook | Not on Free |
| Opus, and Fable | Not on Free |
| Usage credits (buying more capacity) | Not on Free |
| Priority access at high traffic times | Not on Free |
| Sonnet, and Haiku | Included |
| Model training | Opt-out available |
That line is coherent, not arbitrary. The free plan gives you Claude in a chat window. It does not give you Claude inside your editor, your browser, your inbox or your documents, and it does not give you the frontier models. The pricing page's tooltip on the Fable row spells the top end out: "Claude Fable 5 is included on Max plans at 50% of your weekly usage limit. Pro users can access Fable via usage credits."
Projects being paid-only matters more than its position in the list suggests. Anthropic's help centre describes projects as the main way to stretch your allowance: "Content in projects is cached and doesn't count against your limits when reused." The feature that makes a usage limit go further sits behind the paywall, and the plan with the tightest limit is the one that cannot use it.
How much free is: Anthropic will not say, and that is the finding
This is the question every reader arrives with, so here is the honest answer instead of a confident one. There is no published number. We read the pricing page and four Anthropic help-centre articles looking for one, and this is the closest anybody comes, from Anthropic's own getting-started article:
"While using the free Claude plan, there is a session-based usage limit that will reset every five hours. Also, the number of messages you can send will vary based on demand, and we may impose other types of usage limits to ensure fair access to all users. Claude will notify you when you've reached your limit, or if your prompt exceeds the available context window."
Take that apart, because there is a lot in it. The limit is a rolling five-hour session budget, not a daily or monthly allowance. Its size is explicitly not fixed: it "will vary based on demand", so the same behaviour can get you a different amount of Claude on a Tuesday morning than on a Sunday night. And Anthropic reserves the right to add "other types of usage limits" without naming them.
Two further details from the same help centre make the ceiling lower than it looks. Usage is pooled across everything: "your usage of all different Claude product surfaces (claude.ai, Claude Code, Claude Desktop) counts towards the same usage limit". And what you spend depends on things you would not think of as spending, the listed factors being message length, file attachment size, conversation length, tool usage such as research and web search, model choice, effort level and artifact creation. Turning on web search costs you allowance.
If you have seen a specific free-message count quoted somewhere, it did not come from Anthropic. We will not repeat it and we refuse to invent one. You will find your limit by hitting it, and it resets five hours later.
The context window, where two Anthropic pages disagree
The context window is how much text Claude can hold in view at once, and it decides whether the free plan can do your actual job. Anthropic publishes two answers and we cannot reconcile them, so you get both.
On the pricing page, the "Context window" row reads 200k in all four individual columns: Free, Pro, Max 5x and Max 20x. Read alone, that says paying buys more turns and not a longer memory in any one of them. But the help-centre article titled "How large is the context window on paid Claude plans?" says something else: "Claude Opus 5 and Sonnet 5 support a 1M token context window on all paid plans when chatting with Claude", with older models at 500K, and 200K described as what you get "when using a paid Claude plan" outside those models. Every figure in that article is scoped to paid plans, and the free plan is not mentioned in it at all.
- Pricing page, Context window row, printed identically in all four individual columns, Free included200k
- Help centre, Claude Opus 5 and Sonnet 5, stated for paid plans only1M
Two Anthropic pages, read the same day. The help-centre figure is scoped to paid plans throughout and the article never mentions the free plan, so this is a gap between two documents and not a measured gap between two plans.
What the free plan costs you instead of money
Anthropic's consumer terms define your inputs and Claude's outputs together as "Materials", and then say: "We may use Materials to provide, maintain, and improve the Services and to develop other products and services, including training our models, unless you opt out of training through your account settings."
The opt-out is real and it is free: the pricing page's "Model training" row reads "Opt-out" on every individual tier including Free. It is a setting, not a default, so it protects nobody who has not gone and changed it, and nothing switched off later reaches backwards over what you typed before.
Two exceptions survive it, and the terms name both: "Even if you opt out, we will use Materials for model training when: (1) you provide Feedback to us regarding any Materials, or (2) your Materials are flagged for safety review". The first is the thumbs up and thumbs down buttons, and the feedback clause is explicit about what pressing one does. If you rate an output, "we will store the related conversation as part of your Feedback", and "we may use the Feedback however we choose without any obligation or other payment to you". One approving click re-enrols that whole conversation. The second exception happens on Anthropic's side, and you are not told.
The clause that catches people who signed up at work
If you opened your free account with a work email address, it may not stay yours. The terms carry a "Business Domains" clause, and it is the most useful paragraph on this page for anyone who has ever pasted something private into a chat window at their desk.
"If you use an email address owned by your employer or another organization, your Account may be linked to the organization's Anthropic enterprise account, and the organization's administrator may be able to monitor and control the Account, including having access to Materials. We will provide notice to you before linking your Account to an organization's enterprise account. However, if the organization is responsible for notifying you or has already informed you that it may monitor and control your Account, we may not provide additional notice."
Read the last sentence twice, because that is where the promised notice quietly evaporates. You get told, unless your employer has already told you at some point that it may monitor your account, which is a line most people agreed to in a staff handbook skimmed on their first morning. A free account is exactly the sort of thing people open on a whim with whatever address is already signed in. The fix takes thirty seconds: use an address you own.
Do this, in this order
- Sign up with a personal email address. One decision now, and the Business Domains clause can never reach the account. If you have already registered with a work address, assume the conversations in it are readable by your employer and behave accordingly.
- Change the training setting before you type anything sensitive, then stop pressing thumbs up. The opt-out lives in account settings and is available on the free plan, so do it on day one. After that, remember that rating an output stores the conversation as Feedback and hands it over for training regardless. That is the trap here, because it looks like a courtesy and works as a consent.
- Expect the wall, and do not plan around a number. The free limit is session-based, resets every five hours, and varies with demand, and free accounts get no usage meter. The only warning is the notification that you have run out. Do not start anything time-critical on a free account.
- Check the two eligibility conditions before you invest any time. Anthropic publishes a list of countries and regions where Claude.ai is offered, and we counted 184 entries on it on July 27, 2026, with several large ones absent. And the terms are unambiguous on age: "You must be at least 18 years old or the minimum age required to consent to use the Services in your location, whichever is higher." Free is only free where you are allowed to have it.
- Only consider paying once you can name the feature you are missing. Most people upgrade chasing a feeling of restriction instead of a measured one. The honest reasons to pay are on the crossed-out list above: Projects, Research, Opus, Claude Code, or the browser and Microsoft 365 integrations. If none of those is what stopped you, a subscription will not fix it.
The rest of the fine print, since free does not mean unconditional
A free account still sits under the whole contract, and a few clauses land differently when you have paid nothing.
You own the output, and the free plan is no exception. As between you and Anthropic, "you retain any right, title, and interest that you have in the Inputs you submit", and Anthropic assigns to you "all of our right, title, and interest" in Outputs.
Access can be withdrawn without notice. Anthropic "may suspend or terminate your access to the Services (including any Subscriptions) at any time without notice to you" if it believes you have breached the terms, and separately reserves the right to "modify, suspend, or discontinue the Services or your access to the Services, in whole or in part, at any time without notice to you". A paying customer terminated for a non-breach reason gets a pro-rata refund. A free user has nothing to refund, so that remedy does not reach you.
If something goes badly wrong, the cap is $100. The liability limit is the greater of what you paid in the previous six months and $100. On a free account the first figure is zero, so one hundred dollars is the whole of it, in capital letters, for everything.
There is no forced arbitration, which is worth saying out loud. We searched the consumer terms for an arbitration clause and a class-action waiver and found neither. Disputes "will be resolved exclusively in the state or federal courts located in San Francisco, California". That deserves credit in a category where the usual move is a private arbitrator and a short window to escape one. Closing the account, though, is email only: "You may close your Account at any time by contacting us at support@anthropic.com."
If you do eventually pay, the money fine print is set out in full in our Claude guide: all payments are non-refundable, and the subscription renews for a term equal in length to the last one unless you cancel at least 24 hours before it ends.
What we could not verify, and will not guess
Three gaps, stated plainly. No free-plan message count exists in any Anthropic document we could read, so none appears here, and nor does a figure for what one web search costs against the allowance. The free plan's context window on the newest models is unestablished, for the reason set out above. And one ambiguity: the terms say "Use of our Services for evaluation purposes are for your personal, non-commercial use only", while nothing we read calls the Free plan an evaluation and the pricing page lists it as a plan alongside the paid ones. The plain reading is that the restriction does not reach it, but no first-party sentence says so in terms, and we would rather show you the seam.
Common questions
Is Claude really free?
Yes. Free is listed at "$0" and called "Free for everyone", and the payment comparison shows "n/a" for both credit card and billing cycle on that column. No card, no countdown, and no discount code, because Anthropic runs no referral or affiliate program to generate one.
How many messages do you get?
Unpublished. Anthropic says only that the free plan has "a session-based usage limit that will reset every five hours" and that the number of messages "will vary based on demand".
What is missing from the free plan?
Claude Code, Cowork, Design and Science, Projects, Research, the Chrome and Microsoft 365 integrations, usage credits, priority access, and the Opus and Fable models. Free keeps Sonnet, Haiku, web search, memory, connectors, skills, artifacts and file creation with code execution.
Does the free plan train on what I type?
By default yes, and there is a free opt-out in account settings. Two things survive it: anything you rate with a thumbs up or down, and anything flagged for safety review. Business plans get "No model training on your content by default", the opposite default to yours.
Can I use it for work, and is it free everywhere?
Nothing in the terms restricts the Free plan to personal use, and you own the outputs. The work risk runs the other way: an account opened on an employer's email domain can be linked to that organisation, after which an administrator may be able to monitor it. And no, it is not available everywhere. Anthropic publishes a country list, and it does not cover the world.
Is it worth your time?
Yes, clearly, on the free plan. It is $0, it needs no card, no link or code improves it by a cent, and it is one of the few free tiers we have read this year that is a working tool rather than a locked showroom. Anyone selling you a Claude coupon is selling something that does not exist.
The comparison people actually want is with the other one, and it is closer than the marketing on either side suggests. ChatGPT's free plan is also genuinely $0, also trains on you unless you opt out, and also carries a work-email clause that can hand your account to your employer. Three differences are real. OpenAI publishes a context window for its free tier and Anthropic does not. Anthropic sends disputes to a court in San Francisco while OpenAI sends them to arbitration unless you opt out within 30 days. And OpenAI's opt-out has no stated exceptions, where Anthropic's has two, one of them a button you will press by accident.
What we would want a friend to do takes five minutes. Open the account on an address you own, turn the training setting off, stop rating outputs, then use it until the wall arrives and let the wall, and not the marketing, tell you whether to pay. The thing that would improve this page most is a published free-tier limit, and until Anthropic prints one, nobody honest can give you a number.
Second opinions
Other free tiers, checked
Every one of these is read from the company’s own document and dated. See all 59 questions we have answered.