[Verified] Claude's free tier includes web search, memory and MCP. The real wall is 10-20 messages per 5-hour window

Most vendors hold back features to push you toward the paid plan. Claude does the opposite. We checked the official pricing page on 2026-07-31 and again on 2026-09-03, and the free tier carries almost every headline feature. Web search, memory across conversations, file creation, code execution, extended thinking and remote MCP connectors are all switched on for free accounts. What Anthropic does ration is volume. Our angle in this piece: the free plan is not feature-crippled, it is time-crippled, and that changes who should pay.

What free actually gives you

Here is the free feature list as we verified it, not as marketing summarises it:

• Chat on web, iOS, Android and desktop
• Generate code and data visualisations
• Write, edit and build content
• Web search
• Memory across conversations
• Create files and run code
• Desktop extensions that unlock more Claude features
• Connect Slack and Google Workspace
• Integrate any context or tool through remote MCP connectors
• Extended thinking for complex tasks

Two of those need a plain-English gloss. MCP is Model Context Protocol, a standard way to plug outside tools and data into a chat assistant. Extended thinking is a mode where the model works through a problem step by step before answering, which usually helps on harder tasks. Both are normally treated as premium features elsewhere. Here they are free.

The condition on the page is “Free for everyone”. There is no trial clock and no card required. The gotcha line matters, though: commercial use is not permitted on the free tier. If you are billing a client for work Claude helped produce, the free plan is the wrong plan on paper, whatever the feature list says.

The limit that actually bites

The free tier’s ceiling is the message budget. Our check found roughly 10-20 messages of ordinary conversation, and the budget is calculated on a rolling 5-hour window rather than a fixed daily reset. That second detail is the one readers keep getting wrong.

A daily reset means you plan around midnight. A rolling window means every message you send starts its own 5-hour timer, and capacity frees up as old messages age out. In practice, a burst of 15 messages at 9am can leave you locked out until early afternoon, then quietly restored. A slow trickle across the day may never hit the wall at all. The number of messages you get also shifts with what you ask. Long pasted documents and heavy extended-thinking requests eat through the window faster than short questions, which is why we report a range rather than a single figure.

Claude Code, the terminal-based coding tool, is not included at all on free. It starts at Pro.

The paid ladder, with the numbers

All prices exclude tax, and the pricing page states that plans can change at Anthropic’s discretion. Every figure below was verified on 2026-09-03 unless noted.

A seat is one named user on a team plan. Pro adds everything free lacks: more usage, Claude Code, Claude Cowork, Claude Design, Claude Science, unlimited projects for organising chats and files, the Research feature, access to more Claude models, and Claude for Microsoft 365. Max plans add higher output limits, early access to advanced features, and priority access during busy periods. The Max tiers are pure volume multipliers on top of Pro, with the same feature set.

Team plans bundle Claude Code and Cowork, single sign-on, centralised billing, connector management, enterprise search inside your organisation, and a default of not training on your content. You can mix Standard and Premium seats in one workspace. Enterprise adds admin-set usage caps, granular roles, SCIM for identity management, audit logs, a compliance API for monitoring, custom data retention, network-level access control, IP allowlisting, an optional HIPAA-ready configuration, and Claude Security in beta.

One item we flagged internally: the Enterprise price only appeared on the public page on 2026-07-26. Our earlier records had no figure for it. The US$20 per seat is a floor, it is annual-only, and usage is billed separately, so the real cost for a large deployment will be higher than that line suggests.

☀️ AMO, the budget-minded cat

AMO’s question is always the same: is free enough, and if not, what is the cheapest step up?

For a lot of people, free is enough. The feature list is nearly the whole product. You get web search, memory, file creation, code execution, Slack and Google Workspace connectors, and MCP integrations without paying a cent. If your use is a few questions an hour, spread across the day, the rolling 5-hour window will rarely stop you. The 10-20 message range sounds tight, but a message can be long and a good answer can save a dozen follow-ups.

Free stops being enough in two cases. The first is bursty work. If you sit down for a two-hour session and want to iterate on a document, you will hit the window and lose the rest of the afternoon. The second is commercial use, which free does not permit. Either way, the answer is Pro at US$20 a month, or US$200 a year, which the page presents as US$17 a month. Pro is where Claude Code, projects and Research live, so the jump buys tools as well as volume.

AMO’s warning is about the tiers above Pro. Max 5x costs US$100 and Max 20x costs US$200. Those are volume plans, and nothing else. If you are not routinely running out of Pro, they are five to ten times the money for capacity you will not use. Teams should also notice that annual billing drops Team Standard from US$25 to US$20 per seat, which is the same monthly figure as an individual Pro plan with admin controls thrown in.

🌙 PIMI, the performance-minded cat

PIMI cares whether the thing works, and whether the paid plans make it work better.

On capability, the free tier is not a demo. Extended thinking is on, which is the mode that matters for harder reasoning and long documents. Web search is on, so answers can be grounded in current pages rather than training data. Memory across conversations is on, so you are not re-explaining your project every session. Remote MCP connectors are on, which means free users can wire in their own tools. PIMI’s verdict on the free tier as a piece of software: it is the same engine as Pro, throttled.

The throttle is the performance problem. A rolling 5-hour window with a 10-20 message budget punishes exactly the workflow where Claude is strongest, which is long back-and-forth on a big piece of text or code. Extended thinking and pasted documents drain the window faster. So the more demanding the task, the sooner free runs dry.

Paying fixes that in stages. Pro gives more usage plus Claude Code, which is the tool that turns the model into something that reads and edits real repositories. It also unlocks more Claude models, which matters if you want to choose a stronger model for hard problems. Max 5x and Max 20x add higher output limits on all tasks and priority access during high-traffic periods. That priority line is the one performance-minded users should weigh. If you have felt the model slow down or queue at peak hours, the US$100 and US$200 tiers are the only ones that promise to move you ahead. For everyone else, Pro is where the performance gains stop and the volume gains begin.

Verdict: who should use it, who should skip it

Use the free tier if you are an individual doing light, non-commercial work spread through the day. You are getting web search, memory, extended thinking and MCP for nothing, and the 5-hour window will mostly stay out of your way.

Move to Pro at US$20 a month if you work in bursts, need Claude Code, or do anything you bill for. The annual option at US$200 is the best per-month price on the individual ladder.

Consider Max 5x or Max 20x only if you already exhaust Pro or need priority at peak hours. They add volume and queue position, not features.

Teams of 2-150 should look at Team Standard on annual billing at US$20 per seat per month. Enterprise is annual-only from US$20 per seat with usage on top, so budget for more than the headline.

Skip the free tier if your work is commercial or your sessions are long and continuous. The features will tempt you, and the window will stop you.

Written: 2026-09-12 Price last verified: 2026-09-12 Tool last health-checked: 2026-09-12

📎 Sources

All prices and quotas in this article come from AMPM’s verified dataset for Claude (last verified 2026-09-12); primary sources:

• https://claude.ai

🐾 Meet the cats: AMO and PIMI

AMPM is a Taiwan-based AI-tool pricing watchdog. Our motto: ask before you subscribe. Two cats argue every tool from two sides:

• ☀️ AMO — the budget-minded cat. Always asks: is the free tier enough, and is paying actually worth it?
• 🌙 PIMI — the performance-minded cat. Cares about whether it works well, fast, and gets your job done.

When they are done arguing, you know whether to pay. Every number is checked by us, with the verification date at the end.

🐾 Our sister sites also checked

• Travel insurance and flight-delay coverage checked
• Exchange fees and safety checked