Claude Tag: Anthropic's AI Teammate Is Now in Slack

Most AI tools are personal assistants — they wait for you to open an app, type a prompt, and wait for a response. Claude Tag, launched on June 23, 2026, flips that model entirely. Instead of you going to Claude, Claude comes to your team's Slack workspace and works alongside your colleagues the same way a human hire would: with context, memory, tool access, and the ability to take initiative.

If you're studying for the Claude Certified Architect (CCA) exam, Claude Tag is one of the most important product releases to understand right now. It's a real-world implementation of everything the exam tests — agentic deployment, memory management, multi-tool orchestration, and enterprise governance. This guide covers everything.

What Is Claude Tag?

Claude Tag is Anthropic's new Slack integration that transforms Claude from a personal chatbot into a shared AI teammate for your entire team. Available in beta for Claude Enterprise and Claude Team plan customers, it replaces the previous "Claude in Slack" app with a fundamentally different architecture.

Here's the core shift in one sentence: the old Claude Slack app was a 1:1 chat assistant. Claude Tag is a shared, persistent agent that lives in your channels, remembers your team's work, and can act on its own.

When Claude Tag is set up in a channel, any team member can type @Claude followed by a task in plain English. Claude then:

Breaks the task into stages — analyzing what needs to happen before diving in

Uses connected tools — codebases, databases, APIs, external services

Works asynchronously — Claude doesn't block the thread; it works in the background and posts results when done

Shares context across the team — one Claude instance per channel means your colleague can pick up a thread you started without re-explaining anything

This is a meaningful shift in how enterprise AI is deployed. Claude isn't a productivity add-on anymore — it's a team member with a role, permissions, and institutional memory.

Claude Tag vs. the Old Claude in Slack App

Anthropic is replacing the previous Claude Slack integration entirely. Administrators have a 30-day opt-in window for migration, and the old app is being phased out on August 3, 2026. Here's how the two compare:

The upgrade is significant. If you're still using the old integration, the clock is ticking — August 3 is the hard cutoff.

How to Set Up Claude Tag (Admin Guide)

Setting up Claude Tag requires either a Primary Owner or Owner role in your Slack workspace — the Admin role alone is not sufficient.

Step 1: Pair with Slack

Install the Claude Tag app from Anthropic's workspace settings. This creates Claude's identity within your Slack workspace and links it to your Claude Organization plan.

Step 2: Connect Tools and Data

This is where Claude Tag's power comes from. Admins choose which tools Claude can access:

• Codebases — connect your GitHub repos so Claude can read and write code
• Data sources — databases, spreadsheets, internal wikis
• APIs — external services your team uses daily
• MCP servers — for teams already running Model Context Protocol infrastructure

Each connection is scoped deliberately. A channel configured for engineering work gets access to engineering tools. A channel for legal work gets legal data — and nothing else. Claude configured for one domain has no visibility into another.

Step 3: Set Spend Limits

Anthropic built enterprise-grade cost governance directly into Claude Tag. Admins can set:

• Organization-wide token spend limits — a global ceiling across all Claude Tag usage
• Per-channel limits — individual budgets so a single high-volume channel can't drain the whole org's quota

This solves a real pain point: enterprise AI adoption often stalls because finance teams can't predict costs. Claude Tag makes token usage auditable and controllable.

Step 4: Test in a Private Channel

Before rolling Claude Tag out to your whole team, Anthropic recommends testing in a private channel with a small group. Validate that tool connections work, memory is building correctly, and the channel's Claude instance behaves as expected. Once confirmed, expand to public channels.

Core Features That Make Claude Tag Different

Persistent Channel Memory

Claude Tag builds and maintains context by remembering relevant information from the channels it operates in. This memory is:

• Viewable by admins — no black box; you can inspect what Claude knows
• Editable — admins can correct, add, or remove memories
• Deletable — full GDPR-compatible control if needed

This is architecturally important. Claude isn't starting from scratch every conversation. Over time, it learns your team's preferences, recurring workflows, and project context — similar to how a new hire absorbs institutional knowledge.

Asynchronous Task Execution

Unlike a chat assistant that blocks until it responds, Claude Tag works in the background. Tag Claude with a task, then go back to work. When Claude finishes — whether that's 30 seconds or 20 minutes later — it posts results in a Slack thread.

Claude can also plan tasks to complete in the future, making it genuinely proactive rather than purely reactive.

One Claude Per Channel, Shared by the Team

This is one of the most important architectural decisions Anthropic made. There's a single Claude instance per channel — not a separate assistant for each user. That means:

• Continuity — conversations aren't siloed by user; anyone can pick up where someone else left off
• Team awareness — Claude knows the full context of the channel, not just your individual thread
• Accountability — every action is attributable and auditable

Full Audit Logging

Under Organization Settings → Claude Tag → Audit, admins see a complete log of every task Claude has completed, every network call made via Agent Identity, and who requested each action. This level of governance is essential for regulated industries and security-conscious enterprises.

What Claude Tag Means for the CCA Exam

If you're preparing for the Claude Certified Architect certification, Claude Tag is a live case study in several core exam domains.

Agentic Deployment Patterns

Claude Tag implements the multi-tool orchestration pattern that appears throughout the CCA exam. Claude doesn't just answer questions — it decomposes tasks, selects tools, sequences calls, and synthesizes results. Understanding how Claude approaches multi-step workflows (and where it might fail or need guardrails) is essential CCA knowledge.

Memory Architecture

Claude Tag uses persistent external memory — the channel memory that persists between sessions. This is distinct from in-context memory (what fits in the prompt) and from retrieval-augmented generation (fetching from a vector store). The CCA exam tests candidates on when to use each memory pattern. Claude Tag is a practical example of per-deployment persistent memory with admin oversight.

Identity Isolation and Least Privilege

The per-channel identity model — where each Claude instance only has access to the tools assigned to that channel — is a direct implementation of the least-privilege principle for AI agents. The exam asks candidates to design agent systems that minimize risk. Claude Tag's architecture is a reference model.

Token Budget and Governance

Per-channel spend limits reflect the task budget concept introduced with Claude Opus 4.7 and central to Anthropic's enterprise strategy. Architects need to know how to scope, monitor, and constrain agent resource usage — Claude Tag's admin controls are a real-world implementation.

Want to test your knowledge of these patterns? The CCA practice test bank at AI for Anything includes 200+ questions covering agentic deployment, memory, orchestration, and governance.

Who Can Use Claude Tag Right Now?

Claude Tag is available in beta for:

• Claude Enterprise customers (any workspace size)
• Claude Team customers (minimum 5 seats, $30/user/month)

Availability on Claude Pro or individual plans has not been announced. If you're on the old Claude Slack app, watch for the migration prompt from Anthropic — you have until August 3, 2026, before the old app stops working.

Key Takeaways

• Claude Tag launched June 23, 2026 as a beta replacement for the old Claude Slack app — the old app ends August 3, 2026
• It's a shared team agent, not a personal assistant: one Claude per channel with persistent memory, async execution, and tool access
• Setup requires Owner or Primary Owner role in Slack; Admins cannot configure it
• Identity isolation and spend limits give enterprises the governance controls needed for real deployment
• It's a live CCA exam case study covering agentic patterns, memory architecture, least-privilege design, and token budget management

Get Ahead on Claude Agentic Features

Claude Tag is one of the fastest-moving areas in the Claude ecosystem — and it's exactly the type of enterprise deployment the CCA exam was built to certify. The exam tests whether you can design, troubleshoot, and optimize Claude deployments at this level of complexity.

Start practicing with the most up-to-date CCA question bank at AI for Anything — 200+ questions covering agentic AI, memory patterns, MCP integration, and enterprise governance, updated to reflect the June 2026 Claude feature set.

---

Sources: Introducing Claude Tag — Anthropic · TechCrunch: Claude Tag is learning your company · VentureBeat: Anthropic launches Claude Tag · Claude Help Center: What is Claude Tag