Anthropic on Monday released Cowork, a new AI agent capability that extends the power of its highly successful Cloud Code tool to non-technical users — and according to company insiders, the team built the entire feature in about a week and a half, using primarily Cloud Code itself.
This launch marks a major inflection point in the race to deliver practical AI agents to mainstream users, positioning Anthropic to compete not only with OpenAI and Google in conversational AI, but also with Microsoft’s Copilot in the growing market for AI-powered productivity tools.
"Cowork lets you complete non-technical tasks like developers using the cloud to code," The company made the announcement via its official cloud account on X. The feature comes as a research preview that’s available exclusively to Cloud Max customers — Anthropic’s power-user tier costs between $100 and $200 per month — through the macOS desktop application.
Over the past year, the industry narrative has focused on large language models that can write poetry or debug code. With Cowork, Anthropic is betting that the real enterprise value lies in an AI that can open a folder, read a messy stack of receipts, and generate a structured expense report without a human hand.
How developers using coding tools for leisure research inspired Anthropic’s latest product
Cowork’s origins lie in Anthropic’s recent success with the developer community. In late 2024, the company released Cloud Code, a terminal-based tool that allows software engineers to automate rote programming tasks. The tool was a hit, but Anthropic noticed a strange trend: Users were forcing the coding tool to perform non-coding labor.
According to Boris Cherny, an engineer at Anthropic, the company saw users deploy developer tools for an unexpectedly wide variety of tasks.
"Since we launched Cloud Code, we’ve seen people use it for all kinds of non-coding tasks: researching holidays, creating slide decks, cleaning up your email, canceling subscriptions, retrieving wedding photos from a hard drive, monitoring plant growth, controlling your oven," Cherney wrote on X. "These use cases are diverse and surprising – this is because the built-in cloud agent is the best agent, and Opus 4.5 is the best model."
Recognizing this shadow usage, Anthropic effectively removed command-line complexity from its developer tools to create a consumer-friendly interface. Anthropic explained to developers in its blog post announcing the feature. "quickly began using it for almost everything," Who "What inspired us to create Cowork: an easy way for anyone – not just developers – to work the same way with the cloud."
Inside the folder-based architecture that lets the cloud read, edit, and create files on your computer
Unlike a standard chat interface where the user pastes text for analysis, Cowork requires a different level of trust and access. Users specify a specific folder on their local machine that the cloud can access. Within that sandbox, the AI agent can read existing files, modify them, or create entirely new files.
Anthropic offers several illustrative examples: reorganizing a disorganized Downloads folder by sorting and intelligently renaming each file, creating a spreadsheet of expenses from a collection of receipt screenshots, or drafting a report from notes scattered across multiple documents.
"In Cowork, you give the cloud access to a folder on your computer. The cloud can then read, edit, or create files in that folder," The company explained on "Try creating a spreadsheet from a pile of screenshots, or create a first draft from scattered notes."
Architecture depends on what is known as a "Agentic Loop." When a user assigns a task, AI doesn’t just generate a text response. Instead, it formulates a plan, executes steps in parallel, checks its work, and asks for clarification if any obstacles arise. Users can queue up multiple tasks and let the cloud process them simultaneously – a workflow Anthropic describes as feeling. "It’s less like commuting and more like leaving a message for a colleague."
The system is built on Anthropic’s Cloud Agent SDK, meaning it shares the same underlying architecture as Cloud Code. Anthropic notes that colleagues "The cloud can handle many of the same tasks as code, but in a more accessible form for non-coding tasks."
The Recursive Loop Where AI Creates AI: Cloud Code Has Reportedly Written Much About Cloud Cowork
Perhaps the most notable detail about Cowork’s launch is the speed at which the tool was reportedly built – highlighting a recursive feedback loop where AI tools are being used to create better AI tools.
During a livestream hosted by Dan Schipper, Anthropic employee Felix Rieseberg confirmed that the team built Cowork in about a week and a half.
Alex Volkov, who covers AI development, expressed surprise at the timeline: "Holy crap Anthropic created ‘Cowork’ in the last…week and a half?!"
This immediately led to speculation as to how much of Cowork was built by cloud code itself. Simon Smith, EVP of Generative AI at Click Health, put it clearly on X: "Cloud Code wrote all the things about cloud cowork. Can we all agree that we are at least somewhat in a recursive recovery cycle here?"
The implication is profound: Anthropic’s AI coding agent may have contributed significantly to the creation of its own non-tech sibling product. If true, it is one of the most visible examples yet of AI systems being used to accelerate their own development and expansion – a strategy that could widen the gap between AI labs that successfully deploy their own agents internally and those that do not.
Connectors, browser automation, and skills extend Cowork’s reach beyond the local file system
Colleagues do not work in isolation. The feature integrates with Anthropic’s existing ecosystem of connectors – tools that connect the cloud to external information sources and services like Asana, Notion, PayPal, and other supported partners. Users who have configured these connections in the standard cloud interface can take advantage of them within cowork sessions.
Additionally, Cowork can pair with Anthropic’s browser extension, Cloud in Chrome, to perform tasks that require web access. This combination allows the agent to navigate websites, click buttons, fill out forms, and pull information from the Internet while operating from a desktop application.
"Cowork includes a number of innovative UX and security features that we think make the product truly special," Cherney explained while highlighting "a built-in VM [virtual machine] For isolation, out-of-the-box support for browser automation, support for all your claude.ai data connectors, asking you for clarification when unsure."
Anthropic has also introduced an introductory set "Skill" Designed specifically for Cowork, it extends the ability to create documents, presentations and other files in the cloud. These are based on the Skills for Cloud framework, which the company announced in October, which provides specialized instruction sets that the cloud can load for particular types of tasks.
Why Anthropic is warning users that its own AI agent may delete their files
The transition from a chatbot suggesting edits to an agent making edits brings significant risks. An AI that can organize files can, theoretically, delete them.
In a remarkable display of transparency, Anthropic devoted considerable space in its announcement to warning users about the potential dangers of cowork – an unusual approach to a product launch.
The Company clearly acknowledges that the cloud "It can perform potentially destructive actions (such as deleting local files) if instructed to do so." Because Cloud may sometimes misinterpret instructions, Anthropic urges users to provide instructions "very clear guidance" Regarding sensitive operations.
Of greater concern is the risk of prompt injection attacks – a technique where malicious actors embed hidden instructions in the content the cloud may encounter online, potentially allowing the agent to bypass security measures or take harmful action.
"We have built sophisticated protection against rapid injection," Anthropic wrote, "But agent security – that is, the act of securing the real-world activities of the cloud – is still an active area of development in the industry."
The company described these risks as inherent to the current state of AI agent technology rather than unique to Cowork. "These risks aren’t new with cowork, but this may be the first time you’re using more advanced tools that go beyond a simple conversation," Announcement note.
Anthropic’s desktop agent strategy poses a direct challenge to Microsoft Copilot
The launch of Cowork puts Anthropic in direct competition with Microsoft, which has spent years attempting to integrate its Copilot AI into the framework of the Windows operating system, with mixed adoption results.
However, Anthropic’s approach differs in its isolation. By limiting the agent to specific folders and requiring explicit connectors, they are attempting to strike a balance between the usefulness of an OS-level agent and the security of a sandboxed application.
What distinguishes Anthropic’s approach is its bottom-up development. Rather than designing AI assistants and retrofitting agent capabilities, Anthropic first created a powerful coding agent – Cloud Code – and is now abstracting its capabilities to a broader audience. This technological lineage may give Cowork more robust agentic behavior from the start.
Cloud Code has generated significant excitement among developers since its initial launch as a command-line tool in late 2024. The company expanded access with a web interface in October 2025, followed by Slack integration in December. Cowork is the next logical step: bringing the same agentic architecture to users who may never touch a terminal.
Who can access Cowork now, and what’s coming next for Windows and other platforms
For now, Cowork remains only for Cloud Max customers using the macOS desktop application. Users at other subscription tiers – Free, Pro, Team, or Enterprise – can join the waiting list for future access.
Anthropic has signaled clear intentions to expand the reach of the feature. The blog post explicitly mentions plans to add cross-device sync and bring Cowork to Windows as the company learns from the research preview.
Cherney sets expectations appropriately by describing the product "Early and raw, which is how Cloud Code felt when it first launched."
To access Cowork, Mac customers can download or update the Cloud macOS app and click "Colleague" In the sidebar.
The real question of enterprise AI adoption
For technology decision makers, cowork has implications beyond any single product launch. The barriers to AI adoption are changing – now the limiting factor is not model intelligence, but workflow integration and user trust.
Anthropic’s goal, as the company says, is to make working with the cloud feel less like operating a device and more like delegating a task to a coworker. Whether mainstream users are willing to hand over folder access to an AI that might misinterpret their instructions remains an open question.
But the speed of Cowork’s development – a key feature built in ten days, presumably by the company’s own AI – previews a future where the capabilities of these systems grow faster than organizations can evaluate them.
The chatbot has learned to use the file manager. What it learns to use next is anyone’s guess.
<a href