Anthropic launches Claude Tag, integrating collaborative AI coworkers directly into Slack channels

Anthropic has announced the launch of Claude Tag, a new enterprise-focused product designed to integrate its artificial intelligence directly into Salesforce's Slack workspace[1][2]. Unlike standard, individual-use chatbots that operate in isolated, private messaging windows, Claude Tag functions as a persistent, shared colleague within group channels[1][2]. Users can simply type @Claude in any authorized channel to assign complex, asynchronous tasks to the AI, which can then collaborate with multiple team members simultaneously[3][2]. Highlighting the capabilities of this new system, Anthropic revealed that its own product engineering team already relies on an internal version of Claude Tag to generate sixty-five percent of its codebase[3][1]. This striking statistic positions the tool not just as an experimental helper, but as a core engine of the company's internal software development[1].

At the core of Claude Tag is a shift from single-player AI assistants to multiplayer, cooperative agents[1][2]. By establishing a single, unified Claude identity within a Slack channel, the platform allows all members of a team to witness the AI's operations in real time[4][2]. When a team member delegates a task, Claude breaks the request down into logical, multi-stage steps, posting a visible checklist directly in the channel[3][4]. Because the interaction is public, any team member can step in, review what the agent has created, and pick up the thread from where the last person left off without needing to repeat context or re-explain the project history[3][4]. Furthermore, Claude Tag is equipped with persistent memory that continuously follows and contextualizes ongoing conversations over long periods[5][2]. This enables it to handle multi-day projects asynchronously, completing tasks and reporting back with final results even if the original human requester has gone offline[5][6].

The utility of the tool extends beyond reacting to direct queries, as it is engineered with what Anthropic terms ambient proactivity[2][6]. When authorized, the AI agent can quietly monitor information flows within channels, allowing it to autonomously interject to surface relevant cross-channel metrics, triage incoming alerts, or follow up on dormant, unresolved threads[4][2]. While its ability to write code is its most prominent feature, Anthropic reports that this proactive coworker model is rapidly spreading well beyond its engineering department[3][1]. The company’s internal teams use the system to chase down product metrics, organize project updates, analyze support tickets, and diagnose the root causes of complex software bugs[3][4]. Cat Wu, Anthropic’s head of product for Claude Code, noted that the form factor of tagging the AI like a human teammate is exceptionally powerful, noting she uses the integration to bridge external notifications, such as email alerts, directly into internal collaborative threads[2].

Under the hood, Claude Tag is powered by Anthropic's most advanced model, Claude Opus 4.8, and is initially being made available in a beta and research preview for customers on Claude Enterprise and Claude Team plans[5][2]. Because introducing a shared AI agent into a company’s primary communications hub raises significant data security and governance concerns, Anthropic has paired the tool with highly granular administrative controls[5][2]. Team administrators must explicitly grant Claude access to specific Slack channels, codebases, internal data sources, and tools, ensuring that sensitive information—such as legal documents or human resources conversations—remains isolated from unrelated workspaces[3][2]. Administrators can also establish token spending limits and monitor detailed activity logs to maintain tight oversight over operational costs and data access[6]. However, industry observers note that questions remain regarding how long retained channel context is stored and under what conditions it is cleared, representing a critical security hurdle that enterprise IT leaders will need to evaluate before enabling ambient modes[6].

The release of Claude Tag represents a major strategic escalation in the competitive landscape of enterprise technology, shifting the focus of major AI vendors from building the smartest individual models to capturing the organizational-context layer[1][7]. By embedding Claude directly where daily decisions are made and institutional knowledge accumulates, Anthropic is challenging more than just traditional AI rivals[1]. Tech analysts point out that this integration places Anthropic in direct competition with enterprise search and data platform giants, including Microsoft Graph, Glean, Snowflake, and Databricks, all of which are racing to become the central nervous system of modern corporate operations[7]. By successfully anchoring its AI as a permanent participant in Slack conversations, Anthropic aims to lock enterprises into its ecosystem, making its models indispensable not because they are merely accessed, but because they understand the historical context of the entire company[1][8].

Ultimately, Claude Tag marks a milestone in the evolution of workplace automation, transitioning the role of artificial intelligence from a passive personal tool to an active digital peer[1][4]. If Anthropic’s internal metrics are any indication, the future of work may involve a drastic restructuring of human labor, where professional teams spend less time writing base code or triaging data and more time directing and auditing the output of their AI colleagues[1][4]. As this beta rolls out more broadly, the success of Claude Tag in Slack will likely establish the blueprint for how collaborative AI agents are integrated across other major workplace suites[3][5]. For an industry that has spent years searching for concrete, high-impact enterprise use cases for generative models, this new multiplayer agent paradigm offers a compelling look at a highly integrated, automated future[1][4].