What Is Claude Tag? Putting an AI Teammate in Slack

Personal Claude chat is useful—but on a team you quickly hit “whose thread is this?”, “whose permissions?”, and “where do results live?” When you take AI you hand work to into team workflows, Claude Tag is a leading candidate.

Claude Tag is how you @Claude in Slack and drive real work with org tools and shared context. Official help describes it as running under its own identity, remembering channel context, and following up when needed (What is Claude Tag?).

What you’ll learn in this article:

• What Claude Tag is, and how it differs from older Slack integrations
• How to choose among Chat, Cowork, and Claude Code
• The channel handoff flow (following official docs)
• What to verify on permissions, billing, and beta eligibility

Availability (official): Beta on Team and Enterprise plans. Currently available in Slack. Features, migration deadlines, and model names can change—recheck official help before you deploy.

Claude TagClaude Tag

A feature that places Claude in Slack channels like a teammate and lets you hand off tasks with @mentions. In channel work it runs under the org’s agent identity (connections admins grant); results are shared with the team in the thread.

What’s changed

Anthropic’s product announcement (Introducing Claude Tag, 2026-06-23) highlights:

• Multiplayer — Claude in a channel is shared; anyone can see progress and steer mid-run
• Learning (memory) — Channel context accumulates so you don’t re-explain from zero every time (within permission bounds)
• Proactivity — With ambient enabled, it may follow up on stalled threads and similar
• Async — Hand off a task and keep working; long-running and scheduled work are in scope

The announcement also cites internal Anthropic usage such as “65% of product-team code coming from an internal Claude Tag.” Treat that as Anthropic’s own published figure, not a guarantee you’ll see the same results.

How it differs from Chat / Cowork / Code

Summarizing the official contrast (How Claude Tag works):

Lens	Claude Tag	Cowork	Claude Code
Where	Slack channels	claude.ai / Desktop, etc. (personal handoffs)	Terminal / IDE, etc.
Access	Org connections an admin attaches to the channel	Personal OAuth connectors, etc.	Local permissions / repos
Visibility	Shared with channel members	Mostly you	Mostly you
Best for	Shared work the team watches together	Personal research and doc drafting	Implementation in a local repo

Short version: Shared team work → Tag. Personal handoffs → Cowork / Code.

How people use it

Official docs describe these surfaces:

1. Tagging in a channel — hand a task to @Claude
2. DM to Claude — private work with personal connectors (billed to the personal account)
3. Slack AI assistant panel — access from inside Slack’s panel

Channel handoff, in practice

A walkthrough close to the official docs:

1. A member asks something like @Claude summarize open items for launch prep
2. Claude starts in a thread and shows progress with a checklist
3. Another member adds direction in-thread (docs say re-mentioning isn’t required)
4. Results return in the thread (summary, files, sometimes a PR link, etc.)

Work runs in a cloud sandbox for the thread, not on someone’s laptop. When code changes are involved, expected artifacts include PRs via the Claude GitHub App and similar external systems.

Channel vs. DM: billing and permissions (important)

Official help draws this line:

• Channel tags — org identity and admin-attached connections; org billing
• DMs — personal claude.ai connectors; personal account billing

Having your own Drive connected does not automatically mean the channel can use it. Channel reach is based on org-side access granted to that channel. Docs suggest asking in-channel: @Claude what can you access from this channel?

What Owners should decide first

Help states setup is Primary Owner / Owner only. Checklist items from the announcement and help:

1. Pair the Slack workspace with Claude Tag
2. Grant tool / repo access (org → workspace → private-channel hierarchy)
3. Set org / channel spend limits
4. Pilot in a private channel first

Also:

• Channels with guests have separate response and access settings
• Enterprise can restrict Tag use by role
• Audit tooling can review tasks and network calls

The permission model is documented around agent identity (the agent acts as a service account, not a person). The design intent includes keeping memory and tools separated across sales vs. engineering channels.

How teams use it

Good fits

• Launch prep, incident response, and other work multiple people watch in one thread
• Ending “it’s stuck in someone’s local Claude and nobody can take over”
• Research, drafts, and PR creation within admin-approved connections (GitHub, Drive, etc.)

Poor / careful fits

• Wanting to stay on personal Pro alone (Tag is Team/Enterprise beta)
• No Slack, or guest-heavy channels without governance
• Piloting with broad connections attached to a sensitive channel

Even for smaller teams, one private channel + least privilege is realistic. You don’t need to grant wide GitHub access to a company-wide channel on day one.

Pre-launch checklist

1. Are you on Claude Team / Enterprise?
2. Is Slack a real place work happens?
3. Can an Owner spin up a private pilot channel?
4. Attach the smallest connections first (e.g., read-heavy)
5. Write a rule: humans approve sends, production changes, and customer outreach
6. Share internally how personal handoffs (Cowork/Code) differ from team handoffs (Tag)

For the personal “hand work off” foundation, read How to use Claude as AI you hand work to; for knowledge-work handoffs, see Getting started with Claude Cowork. Implementation-heavy personal work pairs with Claude Code coverage as that series expands.

FAQ

Q.Can I use Claude Tag on personal Claude Pro?

Official help positions it as a Team and Enterprise beta. Safest assumption: channel Tag isn’t available on personal Pro alone. Recheck help for the latest.

Q.Is this the same as the old Claude in Slack?

Officially, it’s the next-generation experience in the same place (Slack). Memory, proactive follow-ups, and a distinct identity are among the expansions. Migration requires Owner opt-in.

Q.What’s different about asking in a channel vs. a DM?

Channels use org connections and org billing; DMs use personal connectors and personal billing. Shared work → channel; personal-data-only work → DM—that’s close to the official split.

Q.Can I ask it to write code?

Official docs describe a Claude Code–class engine in the sandbox that can open PRs on GitHub. Repo connections and permissions must be granted to the channel by an admin.

Bottom line

Claude Tag moves Claude past personal chat into handing work to a shared Slack teammate. It’s a Team/Enterprise beta; Owner-led permissions and spend limits are the real setup work.

The practical tip: don’t over-expand the feature org-wide, and keep review ownership with people. Designing personal handoffs (Cowork/Code) separately from team handoffs (Tag) makes the model easier to reproduce.

Sources

1. What is Claude Tag?（Anthropic Help Center）
2. Introducing Claude Tag（Anthropic (2026-06-23)）
3. How Claude Tag works（Anthropic Docs）
4. What is Claude Tag? (Japanese help)（Anthropic Help Center）