Anthropic’s Claude “Cowork” AI Agent Launches on Mobile and Web
Anthropic has released a new “cowork” mode for its Claude AI assistant, enabling the model to act as an autonomous agent that can browse the web, use tools, and complete multi-step tasks. The feature, previously available only in a limited beta, is now rolling out to all Claude users on both mobile and web platforms. It lets Claude independently plan and execute actions such as filling out forms, searching for information, or managing files, while keeping the user in control.
What the Cowork Mode Does
The cowork agent operates by breaking down a user’s request into sub-tasks. It then uses a built-in browser, code interpreter, and other tools to carry out each step. The agent pauses when it needs user input or confirmation, then resumes.
- Autonomous task execution — Claude can navigate websites, extract data, and interact with online forms without manual step-by-step guidance.
- Tool integration — The agent uses a built-in code environment and web browser to perform actions like calculating, summarizing, or fetching live information.
- User control — Users can review the agent’s plan, interrupt at any point, and approve or modify actions before they are executed.
The cowork agent is designed to handle “long, complex, multi-step tasks” that would otherwise require the user to constantly prompt the assistant, according to Anthropic.
Availability and Platform Details
The new agent mode is available immediately to all Claude users on the web and on the Claude mobile app (iOS and Android). It does not require a separate subscription but is part of the existing Claude Pro and Claude for Work plans.
- Web access — users can activate the cowork agent from the Claude interface on desktop browsers.
- Mobile access — the same agent capabilities are available through the Claude mobile app, with an interface optimized for touch.
- No extra cost — the feature is included in current subscription tiers, though usage may be subject to standard rate limits.
How It Differs From Standard Claude
Unlike the standard chat mode, which responds to individual prompts, the cowork agent maintains a persistent plan and executes actions in the background. It can revisit earlier steps, correct errors, and ask clarifying questions.
- Persistent context — the agent remembers the full task history and can adjust its approach based on new information.
- Error handling — if a web page fails to load or a tool returns an unexpected result, the agent can retry or suggest alternatives.
- Transparency — Claude displays its reasoning and intended next steps before making changes, allowing the user to override.
Privacy and Safety Considerations
Anthropic has stated that the cowork agent operates within the same privacy and safety guardrails as standard Claude. User data is not used to train the model unless explicitly opted in. The agent cannot access private files outside of what the user explicitly shares.
Anthropic emphasizes that “cowork is designed to be collaborative, not fully autonomous,” and that users always retain final approval over significant actions.
Why This Matters
The release marks a shift from conversational AI to practical agentic AI that can complete real-world tasks on a user’s behalf. For professionals, this means less time spent on repetitive data entry, research, or form filling. For everyday users, it brings the promise of a digital assistant that can actually “do” things rather than just “say” things.
- Productivity gains — reducing the number of manual steps for online tasks cuts down repetitive work.
- Wider adoption — mobile availability makes agentic AI more accessible to users on the go.
- Competitive pressure — other AI companies are racing to offer similar autonomous capabilities, making this a key differentiator.
Background
Claude’s cowork agent was first announced in a limited beta earlier this year. The gradual rollout allowed Anthropic to refine the agent’s planning and error-recovery logic. The current public release incorporates feedback from thousands of beta testers.
- Beta phase — early testers reported the agent handled about 80% of routine web tasks without human intervention.
- Improvements — the final release includes better handling of dynamic web content and improved task decomposition.
- Future plans — Anthropic has hinted at deeper integration with third-party services and offline capabilities.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.