Claude Computer Use Is Now Live on Mobile — No Setup Needed

On March 25, 2026, Claude added interactive apps to its iOS and Android apps — letting users render live charts, build diagrams, and create shareable visual assets inside a conversation without switching to another tool. Two days earlier, on March 23, computer use arrived as a research preview for Pro and Max subscribers, giving Claude the ability to open files, click buttons, and run software on your behalf. Together, these two releases mark a turning point: Claude is no longer just a text assistant — it is becoming a mobile-native agent that acts, not just answers.

Why does this matter now? Because mobile is where most people spend their time, and until this month, AI agents capable of actual task execution were desktop-first experiences requiring separate software setups. Claude just changed that — entirely through the existing app, no installs required.

Four Major Claude AI Upgrades Shipped in 23 Days

Anthropic shipped four significant capability expansions in March 2026 alone. Here is what went live, and when:

• March 2 — Memory extended to all users, including the free tier (Claude now remembers your preferences and conversation context automatically across every session)
• March 17 — Persistent Cowork threads enabled on iOS and Android for Pro and Max subscribers — manage AI-driven tasks from your phone, not just your laptop
• March 23 — Computer use research preview launched for Pro/Max (Claude can open applications, navigate file systems, and execute multi-step workflows without any manual input)
• March 25 — Interactive apps arrive on iOS and Android (Claude renders charts, diagrams, and visual assets inline in conversation, shareable directly from the app)

That is a 23-day sprint from memory democratization to full agentic mobile capability — a pace that rivals anything OpenAI or Google shipped in Q1 2026. For context, ChatGPT's comparable rollout across the same feature categories took several months longer.

Claude Interactive Apps — When the Answer Is a Visual, Not a Paragraph

The interactive apps feature is best understood as an embedded rendering engine (a mini display layer built directly into the Claude mobile chat interface) that produces visual outputs without requiring any third-party apps. Before this update, asking Claude to visualize data or sketch an architecture diagram gave you text instructions — and then you had to do the work yourself in another tool.

Now Claude generates and displays the visual directly in the conversation. What you can request today:

• Live data charts — bar, line, scatter, pie — built from structured data you provide in the prompt
• Flowcharts and diagrams — process maps, org structures, system architecture sketches rendered as actual images
• Shareable visual assets — outputs you can export and send without leaving the Claude app

For a product manager explaining a roadmap to stakeholders, or a student building a presentation from research notes, this collapses a three-tool workflow — chat to copy to design app — into a single conversation on your phone. The chart is the deliverable, generated and displayed inline.

This also narrows the feature gap with OpenAI's custom GPT visualizations, but Claude's implementation works natively on iOS and Android without requiring a browser session or desktop. Two major platforms (iOS and Android) now have feature parity with the web.

Computer Use — Claude Does the Clicking, You Keep the Thinking

Computer use (the capability that lets Claude control a virtual screen session, navigate file systems, open applications, and click interface elements) arrived as a research preview on March 23, 2026, available to Pro and Max subscribers — with no plugin installation or additional setup required.

"Research preview" is an important qualifier here. Anthropic is being transparent that this feature works and is available for testing, but has not been hardened for production-critical workflows. Start with low-stakes tasks until a stable release is announced.

For everyday and developer workflows, the capability is already substantial. Practical examples of what computer use can handle today:

• Running a Python script in a sandboxed environment (a protected container where code executes without touching your actual files or system) and receiving a plain-language explanation of the output
• Searching a local file system for a document, opening it, and returning a summary — all in one request
• Executing a multi-step developer workflow — installing a package, running tests, reading the result — without manual command-line input
• Navigating web interfaces or filling forms as part of a longer automated sequence

This matches task-automation capabilities that ChatGPT recently extended to business subscribers. The key difference: Claude's version requires no additional software. Open the Claude app on your phone, and the capability is there. Our automation guide covers how to evaluate agentic AI features safely before integrating them into real workflows.

Memory Goes Free — The Move That Undercuts ChatGPT

On March 2, 2026, Anthropic extended Claude's memory feature (the system that learns from your conversation history, identifies recurring preferences and project context, and applies that knowledge automatically to future sessions) to all users — including the free tier. No subscription required.

This is a direct competitive move. OpenAI's ChatGPT memory feature remains behind a paywall — a $20/month Plus plan is required to access it. Google Gemini's memory rollout has been inconsistent. Claude's free-tier memory creates a new baseline for what a no-cost AI assistant can do:

• New Claude users immediately start building a persistent AI profile — remembering their name, project context, and writing style — without entering a credit card
• Returning free users experience continuity instead of repeating context at the start of every session
• The switching cost from ChatGPT grows with every Claude session where personal memory builds

Paired with the 1M token context window (where a "token" is roughly three-quarters of an English word — 1 million tokens equals about 750,000 words, or roughly 7 full novels) now in beta on both Sonnet 4.6 and Opus 4.6, Claude can simultaneously hold a massive working document and a full personal history in the same session. That combination is new territory for free-tier AI tools.

Sonnet 4.6 and Opus 4.6: The Models Powering the Upgrade

The February 2026 model launches — Opus 4.6 (February 5) and Sonnet 4.6 (February 17) — provide the technical foundation for every capability above. Both shipped in beta with the 1M token context window. Sonnet 4.6 received targeted improvements across 3 areas:

• Coding — improved code generation accuracy and multi-file debugging capability
• Computer use integration — tighter pairing between the model and the new agentic task-execution layer on mobile
• Agent planning — better at decomposing complex multi-step instructions into reliable, sequenced execution steps

Haiku 4.5 — the lighter, faster model in the Claude family — also received upgrades in the same cycle, targeting speed-sensitive applications where response latency matters more than raw capability depth. Together, all 3 upgraded models reflect Anthropic's positioning of Claude as an agent platform, not just a chat interface.

Claude AI Limitations to Know Before You Build on These Features

A few constraints are worth flagging clearly:

• Computer use is research preview only — production-critical workflows should wait for a stable release announcement before relying on this feature
• Health data analysis is US-only — geographic restriction for regulatory and privacy reasons; Android users also require Android 14 or later and the Health Connect app
• Context window compaction has exceptions — the feature that enables "infinite" conversation length (by summarizing earlier messages to free space) does not cover all use case types
• Pro/Max required for computer use and Cowork — free-tier users get memory and interactive apps, but agentic task execution requires a paid plan

How to Access Every Claude AI Feature Starting Today

All features are delivered through official Claude apps with no additional downloads required beyond your existing plan:

• Free users — Open Claude on iOS (App Store) or Android (Google Play), start a conversation, and ask for a chart from any data you share. Memory is already active and begins building from this session automatically.
• Pro/Max users — Look for computer use in your Claude settings or start a task prompt asking Claude to "open and run" something on your machine. Start with a low-stakes test to assess reliability before delegating anything critical.
• Excel and PowerPoint users — The Office add-ins now share full conversation context across both applications, enabling cross-document AI assistance from a single Claude chat thread.
• Chrome users — The Claude browser extension (paid plans) now supports long-running workflows and slash command shortcuts for frequently repeated tasks.

Not sure which plan fits your workflow? Our Claude AI features setup guide breaks down free-tier memory activation, interactive apps, and how to run your first computer use session step by step.

If you have not opened Claude's mobile app since late 2025, the product you find today is substantially different. Free memory and interactive apps are live with zero setup required. For Pro and Max users, computer use marks the start of Claude as a tool that performs work — not just describes it. Open the app, try a chart, then hand it your first low-stakes task. The delegation ceiling just moved up entirely.

Related Content — Get Started | Guides | More News