Anthropic drops Claude Tag into Slack, replacing its Claude app and turning chat into agents

Anthropic just took the most consequential swing at enterprise collaboration in years: Claude Tag is now embedded inside Slack, replacing Anthropic's existing Claude in the Slack app. The product launches today in beta for Claude Enterprise and Team customers, and it is designed to act like a persistent, shared teammate that anyone on a team can delegate work to by typing @Claude.

This is not “an assistant you talk to.” It is an AI agent built to sit inside the channel where decisions get made and work gets assigned, and then keep context as the team moves. That is why the headline matters for anyone running operations, security, or vendor strategy. The question is no longer “Can AI help with tasks?” It is “Who owns the institutional brain of the org, and how do you govern it when it can operate asynchronously across your tools?”

Here is how Claude Tag works once an administrator sets it up. An IT or security lead pairs Claude Tag with a Slack workspace, grants it access to specific tools and data sources, sets spending limits, and defines which channels it can operate in. After that, any team member in those channels can tag @Claude with a request like writing a pull request, pulling sales numbers, or running a data analysis. Claude breaks the task into stages, executes using the permitted tools, and posts results back into Slack threads.

Anthropic also built four differentiators that are hard to ignore if you have evaluated other Slack AI integrations. First, it is multiplayer. Within a channel, there is one Claude that everyone can see and join, so the work is not trapped in a private DM or single-user loop. Second, it learns over time by accumulating context from the channel it monitors. Users do not have to re-explain projects from scratch, and (with permission) Claude can pull context from other Slack channels and data sources. Importantly, Anthropic says it will not report from private channels.

Third, Claude Tag takes initiative. With ambient behavior enabled, it can proactively surface relevant information from across the channels it monitors and follow up on threads or tasks that go quiet without resolution. That is a meaningful shift in agency: instead of waiting for prompts, the system monitors the information environment and decides what humans likely need next. Fourth, it works asynchronously. Anthropic says its own teams “now spend much more of our time delegating tasks to many Claudes in parallel,” and Claude can pursue projects over hours or days.

All of that raises a very enterprise-sized question: what happens when an AI agent can do things, not just answer questions? Claude Tag is designed with enterprise-grade isolation and administrative governance at its center. System administrators define separate Claude identities for different uses, scoped to specific channels and tied to specific tools and data access. Everything, including accumulated memories, stays within those boundaries, so a Claude configured for sales work will not share memories or data access with one configured for engineering.

There are also controls that map directly onto what security and compliance teams ask for every year. Admins can set token-spend limits at both the organizational and channel level. They can review a complete log of every action Claude has taken and which user requested each task. For organizations dealing with compliance, audit, or regulatory requirements, that logging and scoping architecture is “table stakes,” especially after many enterprises found the lack of such controls a dealbreaker when evaluating AI collaboration tools.

Claude Tag runs on Claude Opus 4.8, the model Anthropic released less than a month ago. Migration requires an administrator opt-in within 30 days, and Anthropic says it is issuing introductory launch credits to eligible Enterprise and Team organizations. The setup is a four-step process, designed to reduce friction for IT teams already managing sprawling SaaS portfolios: pair with Slack, connect tools, set spend limits, then test in a private channel.

And this arrives right in the middle of the most contested territory in enterprise AI: the Slack channel. Slack is actively positioning itself as an “agentic operating system,” and the biggest AI and enterprise software players have been racing to embed themselves into the communication layer where work actually gets coordinated. Salesforce, which acquired Slack for $27.7 billion in 2021, announced more than 30 new capabilities for Slackbot in March. OpenAI introduced “Workspace Agents” in April for enterprise subscribers. Perplexity launched its enterprise “Computer” agent with direct Slack integration. Cognition's Devin has been built around Slack as a primary interface. Microsoft has brought GitHub Copilot into Teams.

The logic behind all of this is brutal and simple: the average enterprise juggles over 1,000 applications, and employees waste countless hours on context switching, draining productivity by up to 40%. The AI system that becomes the default presence in the place where work is coordinated gains an enormous distribution advantage, and also an enormous data advantage, because the agent absorbs institutional context that is difficult to replicate later.

Claude Tag did not pop out of nowhere. Anthropic integrated Claude with Slack in October 2025 with two-way connectivity, expanded in January 2026 with interactive Claude apps for workplace tools like Slack, Canva, Figma, Box, and Clay, and bundled Claude Code into enterprise plans in August 2025 (described by product lead Scott White as “the most requested feature from our business team and enterprise customers”). In April 2026, it launched Claude Managed Agents, APIs for deploying cloud-hosted AI agents at scale, with early adopters including Notion, Rakuten, Asana, and Sentry. Then came Claude Opus 4.8 in late May, described as a “more effective collaborator” with “sharper judgement,” more honesty about progress, and the ability to work independently for longer than its predecessors.

So the strategic stake for executives is straightforward: Claude Tag turns Slack into an execution surface with memory, monitoring, initiative, and asynchronous work. Whoever controls that surface, controls how tasks are delegated, how context is stored, and how auditability is enforced. If you run an enterprise and you are still treating AI as a sidecar, this is your wake-up call. The next procurement cycle will be less about “which model is smartest” and more about “which system can safely operate where the team already works.”