Anthropic has folded its Cowork scheduled‑task capability into the core Claude product, effectively turning Claude into an always‑on agent for its Pro and Max subscribers. Engineers and operators should note that the feature now runs after a user closes a laptop, lives inside the Claude UI, and is being positioned as a repeat‑use tool rather than a headline launch.
Feature merge and capabilities
The Cowork component, which began shipping on July 7, allows Claude to execute jobs on a cloud computer without user interaction. By merging Cowork into the main Claude app, Anthropic makes the always‑on behavior available directly to paying users, with Team and Free plans announced for the near future. Existing Claude Tag in Slack, active since June, provides a proactive teammate with its own identity and audit trail, and Managed Agents for hosted, long‑running workloads have been available since April.
How Claude’s always‑on agent stacks up
OpenAI’s Dots are always‑on agents that include a dedicated cloud computer and browser, and they integrate with more than 4,000 apps. Dots require a ChatGPT Pro or higher plan, priced between $100 and $500 per month. Meta’s Muse, which runs inside WhatsApp, offers a free tier and paid plans at $20 and $100 per month, plus e‑commerce connectors. Anthropic’s Claude agent stays inside the Claude application or is invoked by tagging in Slack, unlike Muse’s WhatsApp integration. The service identity model differs: Claude Tag creates a channel‑scoped identity, whereas Grok Bot (xAI) shares a single cloud computer and login across a roster. These distinctions affect how credentials are handled and how audit data is collected.
Operational and security considerations
Always‑on agents introduce continuous compute consumption, so cost monitoring is essential, especially when the agent runs after a laptop is closed. The shift to a built‑in Claude UI means that any automation still requires the user to open Claude or interact via Slack, which may affect workflow latency. The channel‑scoped identity used by Claude Tag provides a clearer audit trail but also requires careful permission scoping to avoid over‑privileged access. Past incidents, such as the legal and security complications around OpenClaw, illustrate that rapid feature copying can surface unexpected compliance and credential‑exposure risks.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Teams should evaluate whether Claude’s always‑on agent fits their existing toolchain: if Slack is already a central hub, Claude Tag offers a low‑friction entry point; otherwise, the need to launch the Claude app may add friction. Assess the cost of continuous cloud compute against the $100‑$500 monthly range of comparable agents. Review IAM policies to ensure the channel‑scoped identity only receives the permissions required for scheduled jobs, and enable audit logging to track automated actions. Finally, monitor usage patterns to determine if the agent delivers repeatable value beyond the initial novelty, as retention appears to be the primary success metric for these early‑stage agents.
