Anthropic launched Claude Tag on June 23, 2026. It’s not a chatbot you DM. It’s not a one-off integration. It’s a persistent AI teammate that lives in your Slack channels, accumulates context over time, connects to your tools and data, and can act on tasks while you’re doing other things.

The pitch: tag @Claude in any channel, and it responds with the accumulated knowledge of everything it’s observed in that workspace. It remembers conversations from days ago. It can complete multi-step tasks asynchronously. It connects to your codebases, tools, and internal data.

This is a meaningful evolution from the old “Claude in Slack” integration, which was basically a DM chatbot. Let me explain what’s different and why it matters.

How Claude Tag Works

The core concept: Claude Tag creates a single Claude identity per organization that lives across your Slack channels. When you add Claude to a channel, it doesn’t start fresh. It builds context from everything shared in that space, ongoing conversations, decisions, files, links, and uses that context to provide increasingly relevant responses.

The Key Behaviors

1. Persistent context: Claude remembers conversations across days and weeks. Ask it about a decision made last Tuesday and it knows.
2. Channel awareness: It reads channel messages (in channels it’s been added to) and builds understanding of projects, team dynamics, and ongoing work.
3. Tool connectivity: Admins connect Claude to internal tools, databases, code repositories, and documentation. Claude can pull information from these sources.
4. Async execution: Tag Claude with a complex task and it can work on it in the background, posting updates as it makes progress.
5. Shared identity: Everyone in a channel interacts with the same Claude. It’s collaborative, like a team member everyone talks to, not individual instances.

What “Always-On” Means

Claude Tag doesn’t just respond when tagged. It can:

• Follow up on tasks you’ve assigned
• Notice relevant context in channel conversations
• Proactively surface information when it’s relevant
• Complete multi-step work while you’re away

Anthropic describes this as moving from “on-demand help” to “proactive teammate.” Whether that’s exciting or creepy depends on your perspective about AI in workplace communication.

Who Gets Access

Claude Tag is in beta (Anthropic calls it “research preview”) for:

• Claude Enterprise customers
• Claude Team customers

It’s available in Slack only. No Microsoft Teams, no Discord, no other platform (yet). If your organization uses both Slack and another platform, Claude Tag only covers the Slack side.

What Replaced the Old Integration

Before Claude Tag, there was “Claude in Slack,” a simpler integration that let you:

• DM @Claude for one-off questions
• Tag @Claude in a channel for immediate help
• Use Claude Code in Slack to route coding tasks to coding sessions

Claude Tag replaces all of this with a unified experience. The old integration felt like calling a helpline. Claude Tag feels more like having a colleague who’s always available. The persistent context is the difference.

What Claude Tag Can Do

In conversations:

• Answer questions with channel context (“based on last week’s discussion about the migration…”)
• Summarize threads
• Draft messages, emails, documentation
• Translate between languages
• Analyze shared files and links

With connected tools:

• Pull data from internal databases
• Check CI/CD status from connected build systems
• Look up documentation from connected knowledge bases
• Access code repositories for context on technical decisions

Async tasks:

• Research a topic across connected sources and report back
• Draft a document based on channel discussions
• Prepare meeting summaries and action items
• Complete multi-step coding tasks (via Claude Code integration)

What it can’t do:

• Access channels it hasn’t been added to
• Take actions without explicit permission (no surprise deploys)
• Access data outside connected tools
• Override admin permissions or access controls

The “65% of Its Own Code” Claim

Anthropic mentioned that Claude writes 65% of its own code. This is about Anthropic’s internal development process, not Claude Tag specifically. But it signals how deeply they trust Claude for software tasks, and that confidence shows in Claude Tag’s coding capabilities when connected to a codebase.

If you’re building software and tag @Claude with a bug report or feature request, it can potentially write the fix, submit it for review, and report back in the thread. This is the Claude Code in Slack capability, now unified under the Claude Tag identity.

Privacy and Data Considerations

This is where teams should think carefully:

• Claude Tag reads messages in channels it’s been added to
• It retains context across conversations (that’s the point)
• Connected tools expose data to Claude
• Anthropic’s data retention policies apply

What Anthropic says: Enterprise customers’ data is not used to train models. Conversations with Claude Tag follow the same data handling as other Claude Enterprise interactions.

What to think about: Every message in a Claude-enabled channel is processed by Claude. Sensitive conversations (HR issues, confidential strategy, personal matters) should happen in channels without Claude Tag, or in DMs that don’t include Claude.

How It Compares

Claude Tag isn’t the only AI-in-workplace-chat option:

• ChatGPT in Slack: Per-message, no persistent context, no tool connections beyond basic plugins. See our Claude Tag vs ChatGPT in Slack comparison.
• Microsoft Copilot in Teams: Deep M365 integration, summarization, but more assistant-mode than teammate-mode.
• Custom bots: You can build your own with any model, but you’re handling context management, tool connections, and UX yourself.

Claude Tag’s differentiator is persistence. The other options treat each interaction as mostly independent. Claude Tag treats your workspace as an ongoing relationship.

Pricing

Anthropic hasn’t announced separate pricing for Claude Tag. It’s included with Claude Enterprise and Claude Team subscriptions. The usage likely counts against your organization’s Claude token allocation.

For comparison, the Claude Enterprise plan starts at $60/user/month. Claude Team is $30/user/month. Claude Tag is an included feature, not an add-on.

Should Your Team Use It?

Good fit if:

• Your team already lives in Slack
• You’re on Claude Enterprise or Team
• You have internal tools/data that would benefit from AI access
• You want to reduce context-switching (stay in Slack instead of switching to Claude.ai)
• Your team is comfortable with AI observing channel conversations

Not ideal if:

• You handle highly sensitive information in most channels
• Your team prefers explicit, on-demand AI interaction
• You’re not on Slack (Teams, Discord, etc.)
• You’re uncomfortable with persistent AI context in workplace communication

Getting Started

For admins ready to set up Claude Tag, we’ve written a complete setup guide with configuration steps and best practices.

For a broader perspective on multimodal AI pricing (relevant if you’re evaluating Claude versus other providers), see our multimodal AI APIs price comparison. And if you’re interested in how AI handles document processing tasks that Claude Tag might encounter, check our DeepSeek Vision complete guide.

My Take

Claude Tag is genuinely different from previous AI-in-chat integrations. The persistent context is the killer feature. Previous integrations felt like calling customer support: you had to re-explain your situation every time. Claude Tag feels like a teammate who was in the meeting yesterday.

That said, I’d recommend starting slowly. Add Claude to one or two project channels first. See how the team interacts with it. Build trust (and understanding of its limitations) before rolling it out widely. The “always-on” nature means mistakes are visible to the whole channel, not just the person who asked.

FAQ

Is Claude Tag free for Claude Enterprise customers?

It’s included in Claude Enterprise and Claude Team subscriptions at no additional cost. Usage counts against your organization’s standard Claude token allocation.

Can Claude Tag access private DMs?

No. Claude Tag only has access to channels where an admin has explicitly added it. It cannot read private DMs or channels it hasn’t been invited to.

Does Claude Tag work with Microsoft Teams?

Not currently. Claude Tag is Slack-only at launch. Anthropic hasn’t announced Teams support, but given the enterprise market, it would be a logical future expansion.

How is Claude Tag different from the old Claude in Slack?

Three key differences: 1) Persistent context across days/weeks (old version was per-message). 2) Tool and data connections (old version was text-only). 3) Async task execution (old version only responded synchronously to messages).

Can I control which channels Claude Tag can access?

Yes. Admins control which channels Claude is added to. It only observes and responds in channels where it’s been explicitly granted access. You can remove it from channels at any time.

Does Anthropic use Claude Tag conversations to train models?

No. Anthropic states that Enterprise customer data is not used for model training. This applies to all Claude Tag interactions.