Claude Cowork: Anthropic's AI Agent for Non-Coders, in 10 Days

Anthropic shipped Cowork on Monday, a new agent that extends Claude Code's file-and-browser powers to people who don't write code. According to company insiders, Anthropic's own team built the feature in roughly a week and a half, largely by using Claude Code itself.

What changed

Cowork lets a user hand Claude a folder of files or a browser task and let the agent work through it, much like a developer directs Claude Code through a codebase. "Cowork lets you complete non-technical tasks much like how developers use Claude Code," the company said via its official Claude account on X.

The feature launched as a research preview limited to Claude Max subscribers, Anthropic's power-user tier priced between $100 and $200 a month, and it currently runs only on the macOS desktop app. Users on Free, Pro, Team, or Enterprise plans can join a waitlist.

Anthropic's Cherny described the origin on X: "Since we launched Claude Code, we saw people using it for all sorts of non-coding work: doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions, recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven." Anthropic said developers "quickly began using it for almost everything else," which "prompted us to build Cowork: a simpler way for anyone — not just developers — to work with Claude in the very same way."

An agent built by an agent

The build timeline is the detail that spread fastest. During a livestream hosted by Dan Shipper, Anthropic employee Felix Rieseberg confirmed the team built Cowork in approximately a week and a half. Alex Volkov, who covers AI developments, reacted on X: "Holy shit Anthropic built 'Cowork' in the last... week and a half?!" Simon Smith, EVP of Generative AI at Klick Health, was more pointed: "Claude Code wrote all of Claude Cowork. Can we all agree that we're in at least somewhat of a recursive improvement loop here?"

Cherny also credited the underlying model, saying the diverse, surprising use cases exist "because the underlying Claude Agent is the best agent, and Opus 4.5 is the best model" — a claim from Anthropic's own team, not an independent benchmark result.

Read as strategy rather than as a single feature, Cowork looks like Anthropic testing whether the same agent loop that made Claude Code popular with engineers can carry over to spreadsheets, inboxes, and shared drives — terrain Microsoft's Copilot has been courting inside Office. That reading is analysis, not a claim Anthropic itself has made about competitors.

What Cowork does, and what Anthropic warns about

Cherny said Cowork ships with "a number of novel UX and safety features," including "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 it's unsure."

Anthropic paired that with an unusually direct warning: the company acknowledged Claude "can take potentially destructive actions (such as deleting local files) if it's instructed to," and urged users to give "very clear guidance" around sensitive operations. "We've built sophisticated defenses against prompt injections," Anthropic wrote, "but agent safety — that is, the task of securing Claude's real-world actions — is still an active area of development in the industry."

Cherny set expectations accordingly, calling the product "early and raw, similar to what Claude Code felt like when it first launched."

The open question

A feature that reaches a user's real files after just ten days of development puts a specific tradeoff in front of any team evaluating it: speed against the ability to actually vet what's safe to automate. As the original report put it, "the speed of Cowork's development — a major feature built in ten days, possibly by the company's own AI — previews a future where the capabilities of these systems compound faster than organizations can evaluate them." Anthropic's own safety notice reinforces that point: the company says agent security is still unsolved industry-wide, even as it hands Cowork access to local files and browsers. Teams considering Cowork for non-technical staff should weigh that gap against the productivity case before granting broad file access.

What to watch

• Whether Anthropic expands Cowork past macOS and the Claude Max tier — Windows support and lower-tier access would signal a move from preview to product.
• Whether prompt-injection or destructive-action incidents surface as Cowork reaches more non-developer users, given Anthropic's own caveat that agent safety remains unsolved industry-wide.
• Whether Microsoft, Google, or OpenAI answer with a comparable file-and-browser agent built for non-technical users rather than developers.