Anthropic has moved its AI out of the private chat window and into the channel where teams already work. On 23 June it launched Claude Tag, a beta that adds Claude to Slack as a shared member any colleague can summon by typing @Claude. Tag it into a thread with a request and it breaks the task into stages, works through them using the tools and data an administrator has connected, and posts the result back in the thread for everyone to see. Anthropic describes the product as an evolution of its agentic coding tool, Claude Code, and it runs on the company's Opus 4.8 model.

The shift matters because it changes who an AI assistant serves. Earlier integrations gave each person a private Claude tied to their own account; Claude Tag gives a whole channel one shared agent with a single identity, persistent memory and admin-governed access. Slack's general manager, Rob Seaman, characterised the move as making AI "multiplayer". For marketing teams that run campaigns, research and reporting through Slack, the practical effect is that the same agent can be delegated to, corrected and built on by anyone in the room — closer to a teammate than a tool each person prompts alone.

What is Claude Tag, and how does it work in Slack?

Claude Tag is a beta, available to Claude Enterprise and Team customers, that brings Claude into Slack through channel tagging, direct messages and an assistant panel. In a channel, everyone interacts with the same Claude, so one person can pick up a task another started without re-explaining the context. The agent carries four behaviours beyond a standard chatbot: it is multiplayer, shared across a channel; it learns over time by following the conversations and data sources it is granted; it can take initiative, flagging relevant updates and chasing stalled threads when an "ambient" mode is switched on; and it works asynchronously, pursuing a task — or scheduling future ones — over hours or days while colleagues do other work.

Cat Wu, Anthropic's head of product for Claude Code and Cowork, told Fortune that earlier single-user formats made Claude feel like a solo tool, whereas tagging it the way you would a colleague is what makes the channel version distinct. Anthropic says the pattern is already entrenched internally: it claims 65% of its product team's code is now created by an internal version of Claude Tag, with the same approach used to chase product metrics, work through support tickets and trace bugs. That figure is the company's own account of its internal use and has not been independently verified.

How is access and spend controlled?

Access and spend are controlled by administrators before anyone starts tagging. System owners specify which tools, data sources and channels each Claude can reach, creating separate scoped identities for different uses — so a Claude set up for sales work holds no engineering access and passes none of its memory across. Work in a channel runs under the organisation's identity and is billed to the organisation, while direct messages use a person's own connected tools and are billed to their account. Administrators can view an audit log of every task Claude has run and who requested it, with channel posts, code commits and pull requests traceable back to the thread that started them.

The governance detail is aimed squarely at the worries that have slowed enterprise AI adoption: uncontrolled access to sensitive data and the risk of it leaking across teams. Administrators can set hard spend caps for the organisation and for individual channels, with alerts at 75% and 95% of any limit; work that would exceed a cap is declined rather than silently cut short, and a blocked user can request more without leaving Slack. The consumption-based pricing means cost scales with usage rather than headcount — a model that rewards tight scoping and gives finance teams a ceiling to set, echoing the spend limits some companies have already imposed on AI tools.

What does it change for marketing teams?

For marketing teams, Claude Tag relocates AI work from private windows into the shared channel, where it can be supervised in the open. A team could tag @Claude to pull campaign metrics, draft from a brief against connected sources, triage inbound requests or follow up on a quiet thread — with the whole exchange visible, auditable and open to anyone to steer. The trade-off is that an always-on, proactive agent acting under an organisation identity needs careful scoping; the controls exist, but the responsibility for setting them sits with administrators, not the marketers tagging Claude in.

The launch also carries a deadline. Claude Tag replaces the existing Claude in Slack app, and administrators have a 30-day window to opt in before workspaces are switched over automatically on 3 August; Anthropic is offering launch credits to eligible organisations to trial it. The company has framed Slack as a starting point and says it plans to extend the capability to other platforms, though it has not given a timeline or named which.

Claude Tag arrives as Anthropic pushes harder into the enterprise market, where outlets have reported the company overtaking rivals in business adoption. Whether a single shared agent that schedules its own work and acts proactively proves manageable across a busy organisation — rather than adding noise — is the question the beta period will answer.