Every Claude feature, and which plan you need for it

Projects, Memory, Artifacts, Research, Skills, Connectors, Cowork, Claude Code, Chrome and Excel - what each feature does and which Claude plan unlocks it.

Claude's feature set splits along one line: what you get for $0, and what needs a paid plan. As of August 2026 the Free plan includes web search, memory, file creation, code execution, Skills, extended thinking, voice mode, Artifacts and up to five Projects. Everything agentic - Claude Code, Cowork, Research, Design, Science, Chrome and Excel - starts at Pro. This page is the row-by-row matrix.

Every cell below was checked against Anthropic's own pricing page and help centre on 21 August 2026. This is an independent reference site; it does not sell plans and links out to Anthropic for anything account-related.

#Correcting the record: Claude has mobile apps

Two related myths are worth killing at the same time. The Free plan is not browser-only - the mobile and desktop apps sign in with a free account perfectly happily. And there is no separate purchase for Claude Code: it is bundled into Pro, Max, Team and Enterprise at no extra charge.

#The full feature matrix

Twenty-three rows, five plans. Max covers both 5× and 20× unless a row says otherwise.

Three cells deserve a footnote. SCIM on Team is genuinely ambiguous: Anthropic's pricing page lists it under Enterprise, while the SSO help article mentions SCIM without restricting it - hence “not documented”. Memory on Team and Enterprise was described as a pending rollout when checked. And Artifacts carries a documented conflict: the help centre lists them on all five plans, while the Claude Code feature table gates them to Pro/Max/Team with Enterprise admin approval. Those almost certainly describe chat artifacts and Claude Code artifacts respectively, but Anthropic has not said so.

#What each chat feature actually does

#Projects, Memory and chat search

Projects bundle chats with a shared knowledge base and custom instructions. The Free cap of five Projects without retrieval is the meaningful restriction - a free Project holds files, but a very large knowledge base will not be searched intelligently. Memory builds standing notes about you across conversations. Chat search retrieves from past conversations on demand and is the single most-missed paid feature, because on Free your only way back to an old chat is scrolling. All three are walked through hands-on in the guide to using claude.ai.

#Research versus web search

Web search is one query and a cited answer, on every plan. Research is an agent: it runs chains of searches that build on each other, decides what to look at next, pulls in connected Gmail, Calendar and Docs where authorised, and returns a report with checkable citations. It requires web search to be enabled, is paid-only, and burns usage considerably faster than ordinary chat.

#Connectors, MCP and Skills

Connectors are how Claude touches your systems, built on the Model Context Protocol - the open standard Anthropic donated to the Linux Foundation's Agentic AI Foundation in December 2025. Anthropic's own directory lists over a hundred, and its documentation names Notion, Asana, Sentry, Stripe, HubSpot, GitHub, Airtable and Zoom among them. Free accounts get the directory and Google Workspace connectors but only one custom remote MCP server. If you are building rather than connecting, start with MCP and Claude agents.

Skills are dynamically loaded folders of instructions and scripts. Anthropic ships document-creation Skills that trigger automatically; partners including Notion, Figma and Atlassian publish their own; and you can upload custom Skills on any plan, Free included.

#The agentic products: Claude Code, Cowork, Design, Science

Agent Teams deserves clarification because it is often listed as a plan feature. It is not sold separately: it is a Claude Code capability in which peer agents hold independent context windows and message each other under a structured protocol, as opposed to subagents which return a summary to the main thread. Its availability simply follows Claude Code's, and Anthropic publishes no per-plan gating for it. In cloud sessions it is off by default.

Claude in Chrome is a browsing agent extension for desktop Chrome on paid plans; Anthropic's own documentation warns that letting Claude act directly on websites remains risky despite safety classifiers, and Team and Enterprise admins can allowlist or blocklist sites. Claude for Microsoft 365 covers Excel, PowerPoint and Word generally, with Outlook in beta, on all paid plans - with one caveat worth reading twice: Claude for Excel does not inherit an organisation's custom data retention settings and its activity does not appear in Enterprise audit logs or the Compliance API.

#Which features change the model itself?

Two rows in the matrix are about capability rather than tooling.

Model selection is a paid feature. Free users get whatever model Anthropic assigns and cannot switch; Anthropic does not publish which model that is. Paid users choose between Sonnet 5, Opus 5 and older releases, with Fable 5 reserved for Max and premium seats, or available to Pro through usage credits. The full lineup with specifications is on the models page.

The 1M-token context window is a paid-plan feature in chat, and it is model-dependent even there: 1M on Opus 5 and Sonnet 5, 500K on the 4.6–4.8 generation, 200K on everything older. Note that a large window is not the same as reliable recall - on independent long-context retrieval benchmarks Claude trails Gemini by a wide margin, which is worth knowing before you paste a whole codebase in.

#How usage limits interact with all of this

Every feature above draws from the same allocation. There is no separate Design budget, no separate Cowork budget, no separate Claude Code budget.

What consumes the allocation fastest: long conversations, large attachments, Research, Cowork in Auto mode, and high effort levels. What does not: automatic context compaction on paid plans, and re-referencing cached project knowledge. Pro and Max subscribers can buy extra usage credits billed at API rates, capped at $2,000 of redemption per day.

#Frequently asked questions