How Claude Tag Works Inside Your Slack Channels

Claude Tag lets your whole Slack team tag @Claude into a thread, track its progress, and steer it together in real time.

---

Your team's best ideas die in Slack threads. Someone asks a question, three people half-answer it, and by Friday nobody remembers what got decided. Now imagine adding an AI assistant to the mix, except it only remembers what you personally told it, in a private chat nobody else can see.

That's the problem with most AI tools bolted onto team chat. They're single-player. One person prompts, one person reads the answer, and the rest of the channel never sees what happened or why.

Claude Tag fixes that by making Claude an actual member of the channel. Anyone can tag it in, anyone can watch it work, and anyone can jump in to redirect it, mid-task, without starting over.

What Is Claude Tag?

Claude Tag is Claude, living inside your team's Slack channels. An admin connects it to the tools your team already uses (Drive, GitHub, your ticketing system, whatever applies—see our step-by-step setup guide to get started). Once that's set up, anyone in the channel can tag @Claude with a task and it gets to work.

No setup per person. No separate login. You type @Claude followed by what you need, and a working session starts right there in the thread (to learn how to structure these requests, check out our guide on tagging Claude with closeable tasks).

@Claude where are we on launch prep? Pull together what's still open from this channel.

That single line is the whole interaction needed to kick off a task.

How a Session Actually Works

Here's the loop every Claude Tag session follows, from the moment you tag it to the moment it disappears:

1. Tag @Claude with a task
        ↓
2. A sandbox builds (a private workspace for this thread)
        ↓
3. Claude works through the task using the channel's tools
        ↓
4. The result posts back in the thread
        ↓
5. The sandbox shuts down until the next message

Step 2 and 3 happen out of sight. While they're running, you'll just see an "is thinking…" line. For bigger tasks, that line turns into a live checklist that updates as Claude finishes each step:

Done: Read 14 open threads
Done: Cross-checked the launch plan in Drive
Done: Listed who each item is waiting on
Done: Drafted the status summary

Slack doesn't send a notification every time that checklist updates, so a thread can look quiet for several minutes while Claude is actually making steady progress. If something checked off since you last looked, it's working. If it hits a real problem, it'll say so in a reply, not just go silent.

Anyone in the Channel Can Steer It

This is the part that makes Claude Tag different from a normal AI chat. Once a session starts, it isn't just the original asker's task anymore. It belongs to the whole channel.

Say Jordan tags @Claude to summarize open launch items. A teammate, Sam, can jump into that same thread and add:

fold in the vendor quotes from last week's thread too

No need to re-tag @Claude. No need to start a new conversation. Claude reads that reply and folds it into the work already in progress.

A couple of things to know about this:

• Editing or deleting your message afterward doesn't undo what Claude already read. There's no rewind.
• If you need to back out of an instruction, say so in a new reply, or start a fresh thread.

Channels vs. Direct Messages

Where you talk to Claude changes whose access it's using and who can see the result.

	Channel	Direct Message (DM)
Access used	Tools an admin connected to that channel	Your own personal connectors
Who it acts as	Claude's own service account	You
Who sees the work	Everyone in the channel	Just you
Best for	Shared team work	Personal, one-off tasks

In short: use a channel when the team should see and build on the result. Use a DM when it's just for you.

Claude Tag vs. Cowork vs. Claude Code

Anthropic has a few ways to delegate work to Claude. They overlap in ability but differ in where the work happens and who gets to watch.

	Claude Tag	Cowork	Claude Code
Where	Slack channels	claude.ai chat	Terminal or IDE
Access used	Team's shared service account	Your personal connectors	Your local files and credentials
Visible to	Everyone in the channel	Just you	Just you
Best for	Shared, team-visible work	Personal research and drafting	Hands-on coding in your own project

If it's team work the group should see, use Claude Tag. If it's personal, use Cowork or Claude Code.

What Claude Tag Can See and Remember

Even without any extra tools connected, every Claude Tag session starts with a baseline:

• It can read its own thread, plus the channel's pinned items and history
• It can search the workspace for relevant content
• It can write and run code inside its sandbox (this is how it turns a CSV into a chart, or a long thread into a clean summary)

When you tag @Claude partway into an existing thread, it reads up to 50 messages from the start of that thread. In a very long thread, the newest messages right before your tag might fall outside that window, so it's worth restating anything important.

Memory works differently depending on the channel type:

Channel type	Where memory is stored	Who can see it
Public channel	Shared workspace memory	Any channel in the workspace
Private channel	That channel's own memory	Only that channel

So if Claude learns something useful in a public #launch-week channel, it can use that same fact later in #gtm-west. Private channels don't share outward like that.

You can always check what it knows by asking directly:

@Claude what do you remember about this channel?

Anyone in the channel can correct or remove what it remembers.

Checking What It Can Access

Because access is set per channel (not per person), the way to know what Claude can reach isn't to guess from your own permissions. Just ask:

@Claude what can you access from this channel?

If it says no, that channel simply hasn't been granted that access yet. Another channel might have it, and an admin can extend it.

One-Off Tasks vs. Scheduled Tasks

A session can start two ways:

1. Someone tags it manually for a one-time task. The sandbox builds, does the work, and shuts down when finished.
2. A schedule or trigger starts it automatically (a recurring digest, a repository event, a watch on a channel). Same loop, same channel access, just no human typing the request that time.

What Survives Between Replies

The Slack thread itself never goes away. But the sandbox behind it (the actual workspace where Claude does the work) gets released whenever the thread goes quiet, and rebuilds fresh the next time someone replies.

	Survives a quiet period?
The conversation and its context	Yes
Channel memory	Yes
Anything pushed to GitHub, Drive, etc.	Yes
Files that only exist inside the sandbox	No, Claude can recreate them if asked

For longer tasks, it's worth asking Claude to push branches or post drafts as it goes, rather than waiting until the very end. That way the work is saved somewhere permanent even if the sandbox resets.

Quick Recap

• @Claude plus your task in a channel starts a session
• A checklist (or a direct reply) shows progress; silence usually just means it's working
• Anyone in the channel can redirect the task by replying, no re-tagging needed
• Access and memory follow the channel, not the person
• DMs run on your own personal connectors instead of the team's

Q&A

1. What exactly is Claude Tag?

It's Claude embedded directly in your team's Slack channels, set up by an admin, that anyone in the channel can delegate tasks to by typing @Claude.

2. Do I need to set anything up to use it?

No. As long as an admin has added Claude to your channel and connected the relevant tools, you just type @Claude and your request (see the step-by-step setup guide for more details).

3. Can someone else take over a task I started?

Yes. Once a session is running in a thread, anyone in that channel can reply to add instructions, correct a mistake, or steer it, without re-tagging @Claude.

4. Why does the thread sometimes look stuck?

Slack doesn't notify you when Claude edits its checklist. A quiet thread for a few minutes usually just means it's still working through the steps.

5. What's the difference between using Claude Tag in a channel versus DMing it?

In a channel, it uses the team's shared, admin-connected tools and everyone can see the result. In a DM, it uses your own personal connectors and only you see it.

6. How is this different from Claude Code or Cowork?

Claude Code and Cowork are single-player; the work is just for you. Claude Tag is built for a whole channel to see and steer together.

7. Can I undo something I told Claude after I sent it?

Not by editing or deleting your message. Claude already processed what you sent. You'd need to send a follow-up correcting it, or start a new thread.

8. Does Claude remember things between conversations?

Yes, but scope matters. Public channels share memory across the whole workspace. Private channels keep their own separate memory.

9. How do I know what tools or data Claude can access in my channel?

Ask it directly: @Claude what can you access from this channel? It will list exactly what's been granted to that channel.

10. Can Claude Tag run tasks on a schedule, without anyone tagging it?

Yes. It can run as a recurring digest, respond to a repository event, or watch a channel, using the same access a manually tagged request would get.

References

• How Claude Tag works