Anthropic officially launched Claude Tag on June 23, 2026, introducing a persistent AI agent that lives inside Slack channels and retains enterprise memory — a move that replaces the company’s earlier Claude for Slack integration and signals a deeper push into collaborative AI for businesses.

Available as a beta for Claude Enterprise and Team customers, Claude Tag represents a significant evolution from simple chatbot interactions. By tagging @Claude in any channel, teams can summon an AI assistant that not only responds in-thread but also remembers context across conversations, learns from shared documents, and stays present in the channel without needing repeated invitations. This persistent presence, combined with shared memory, allows the agent to function as a collective team resource rather than a personal assistant.

What Exactly Is Claude Tag?

At its core, Claude Tag is an always-on AI agent embedded within Slack channels. Unlike typical bots that require direct messages or explicit mentions to engage, Claude Tag remains active in designated channels, listening to ongoing discussions and building a contextual understanding of the team’s work. When a user types @Claude followed by a query, the agent draws on this stored context — from past conversations, shared files, and even decision history — to provide informed, relevant responses.

The persistent nature is key. Once added to a channel by an admin, Claude Tag doesn’t need to be re-invited for each new question. It maintains a thread of memory, so if a team discusses a project’s status on Monday, the agent can recall that context on Wednesday when asked for an update. That memory is shared across the channel, meaning any member can benefit from the accumulated knowledge without re-explaining basics.

Anthropic has designed the experience to feel like having an expert team member who never sleeps. The agent can answer complex questions, summarize long threads, draft documents, analyze spreadsheet data, and even generate code snippets based on the team’s style and conventions. It integrates natively with Slack’s block kit for rich responses, including buttons, menus, and interactive elements that streamline workflows.

How Claude Tag’s Shared Memory Works

Shared, persistent memory is the headline feature that differentiates Claude Tag from previous Slack AI bots. Early implementations of Claude for Slack were stateless — each query was independent, and the bot had no recollection of past interactions. Claude Tag flips that model.

When an admin activates Claude Tag in a channel, the agent immediately begins indexing the channel’s history, files, and pinned items. It builds a knowledge graph that captures project details, terminology, key decisions, and even team roles. This graph is continuously updated as new messages arrive. Critically, all members of the channel can access this shared memory, so a new team member can ask, “What did we decide about the Q3 budget?” and get an accurate, sourced summary without digging through weeks of messages.

Anthropic emphasizes that memory is not a black box. Users can ask Claude Tag what it remembers, and the agent will list the facts it has absorbed. Admins can also prune or reset memory for specific channels, ensuring that sensitive or outdated information doesn’t linger. The memory operates within the strict data boundaries set by the enterprise’s governance policies, which we’ll explore next.

Built for the Enterprise: Governance and Administrative Controls

For IT administrators, the prospect of an AI absorbing sensitive corporate conversations raises immediate governance concerns. Anthropic addresses this with a comprehensive set of controls tailored for Enterprise and Team plans.

First, Claude Tag is opt-in at the channel level. Admins decide which channels get the agent, and they can limit its scope to specific projects or teams. Once added, the agent respects Slack’s existing permissions: it only sees channels it’s been explicitly invited to, and it cannot access private channels or direct messages without authorization.

Second, data retention policies are granular. Organizations can configure how long Claude Tag retains conversational context — from never to indefinitely — based on compliance requirements. For regulated industries, such as finance or healthcare, this is non-negotiable. Anthropic provides tools to audit and export agent interactions, ensuring that companies can demonstrate compliance with regulations like GDPR, HIPAA, or SOC 2.

Third, role-based access ensures that only specific user groups can query certain types of information. For example, a legal department might allow Claude Tag to recall contract terms but restrict access to privileged attorney-client conversations. These rules are managed through an admin dashboard that integrates with enterprise identity providers.

Anthropic also bakes in content safety filters. The agent refuses to generate disallowed content, and its outputs can be reviewed by moderators before they appear in channels if the organization requires that extra layer of approval. These governance features acknowledge that enterprise AI must be both powerful and predictable.

Replacing the Old Claude for Slack Integration

The launch of Claude Tag effectively sunsets the older Claude for Slack bot that had been available for general workspaces. That earlier integration functioned as a simple conversational interface: you could DM Claude or mention it to get answers, summarize threads, or draft messages, but it lacked memory, persistence, and administration tools.

Existing customers on Enterprise or Team plans will be migrated to Claude Tag during the beta period. Anthropic says the transition will be seamless, with the new agent inheriting existing channel connections. For users accustomed to the old bot, the most immediate difference will be the persistent presence and contextual awareness. No more re-asking the same questions every day; Claude Tag will already know the background.

The company hasn’t set a final date for retiring the older integration, but it’s clear the future lies with the more capable, memory-equipped agent.

Competing with Microsoft Copilot and Others in the Enterprise AI Race

Anthropic’s move puts it in direct competition with Microsoft 365 Copilot, Google Gemini for Workspace, and Salesforce Einstein GPT — all of which embed AI into productivity platforms. While those tools enjoy deep ties to their respective ecosystems, Claude Tag bets on Slack’s position as a hub for cross-platform team collaboration. Many enterprises, especially in tech, media, and professional services, use Slack alongside Office 365 or Google Workspace. Claude Tag doesn’t require organizations to commit to a single productivity suite; it simply slots into the communication tool they already use.

Microsoft Copilot, for instance, is tightly woven into Teams, Word, Excel, and Outlook. Its advantage is the ability to analyze data across those apps, but that also locks users into the Microsoft ecosystem. Claude Tag, by contrast, is platform-agnostic within Slack: it can work with files from Google Drive, Dropbox, OneDrive, and other linked services. Anthropic is positioning Claude as the “Switzerland” of enterprise AI — capable but not prescriptive about where data lives.

Another differentiator is the shared memory angle. Both Microsoft and Google offer some memory features in their AI assistants, but these are largely tied to individual users. Claude Tag’s channel-level, team-wide memory is a novel approach that could appeal to project-oriented organizations.

Potential Challenges and Unanswered Questions

No beta launch is without concerns. The most pressing issue for Claude Tag is data privacy. Even with robust governance, some enterprises may balk at an AI that “remembers” everything said in a channel. Anthropic must prove that its memory is secure, auditable, and easily reversible. Early feedback from beta testers will be crucial in refining these controls.

There’s also the question of accuracy. Like all large language models, Claude can hallucinate — fabricating facts with confident delivery. In a shared team environment, one incorrect memory could misinform an entire channel. Anthropic says it has implemented enhanced grounding techniques that lean heavily on the actual channel content, but the ultimate test will come when users stress-test the agent.

Cost is another variable. Anthropic hasn’t disclosed pricing for Claude Tag beyond the existing Enterprise and Team subscription tiers. It’s possible that heavy usage of the persistent agent will incur additional API or compute fees, which could complicate budgeting for large deployments.

Lastly, there’s the learning curve. Admins will need to understand a new set of governance dashboards, and team members must adapt to having an AI participant that never leaves. Change management may prove to be the soft underbelly of this technical advancement.

What This Means for Windows Enterprise Users

While Claude Tag runs entirely within Slack, it holds relevance for Windows-based enterprises that rely on the Windows ecosystem for daily operations. Many large organizations standardize on Windows 10 or 11 laptops, and Slack’s desktop client for Windows is one of the most installed productivity apps. The ability to summon Claude Tag directly from a channel — without leaving the Slack workspace — aligns with how knowledge workers on Windows already operate.

For CIOs managing Windows fleets, the governance controls in Claude Tag complement existing Microsoft Intune and Azure Active Directory policies. Because the agent respects Slack’s data residency options and integrates with standard identity providers, security teams can fold Claude Tag into their existing compliance frameworks without major upheaval.

Should the beta succeed, it wouldn’t be surprising to see Anthropic develop deeper integrations with Windows features, such as integration with Microsoft Graph for unified search or even a Windows widget for quick access to channel memories. For now, the primary benefit for Windows users is a single, intelligent agent that reduces the need to context-switch between Slack, email, and other tools.

Conclusion and Looking Ahead

Anthropic’s launch of Claude Tag marks a strategic pivot from AI as a personal assistant to AI as a collaborative team member — one that remembers, learns, and adheres to strict governance rules. The June 23 beta opens the door for Enterprise and Team customers to experiment with a new paradigm of shared intelligence, but the real test will be whether organizations can trust an AI to be both omniscient and discreet.

The competitive landscape is heating up, with Microsoft, Google, and Salesforce all vying for a slice of the enterprise AI spend. Claude Tag’s Slack-native approach avoids the vendor lock-in that rivals impose, which could resonate with businesses that value flexibility. As the beta unfolds over the coming months, expect iterations on memory granularity, administrative tooling, and integration depth. For Slack-centric teams, the AI observability and recall that Claude Tag promises might be the productivity boost they didn’t know they needed — but only if they can tame the governance dragon.