Big Idea

The argument in one line.

Claude Tag turns AI from a solo chat tool into a shared, proactive team agent that lives in Slack and acts without being asked — and that architectural shift, not the feature list, is what makes it significant.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You manage a team on Slack and want to delegate research, drafts, or triage tasks without switching tools.
You already use Claude for individual work and want to understand how to extend that to your whole org.
You are on a Claude Enterprise or Team plan and have not yet activated the Slack integration.
You are evaluating whether AI-in-Slack is worth the extra plan cost for your team.

SKIP IF…

You are a solo operator with no team — the multiplayer architecture adds nothing for single-user setups.
Your organization uses Microsoft Teams or another chat platform — Claude Tag is Slack-only at launch.

TL;DR

The full version, fast.

Claude Tag installs Claude as a shared team member inside Slack, giving every person in a channel access to the same AI with accumulated channel context. The standout feature is ambient mode: Claude monitors connected tools and channels and proactively surfaces things the team needs to know before anyone asks. Privacy is controlled at the channel level by admins, with separate memory scopes so sales and engineering do not cross-contaminate. Setup requires a Claude Team or Enterprise plan, takes about five minutes, and works with any app that has a Claude connector or a Zapier MCP bridge.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 02:37

01 · What Is Claude Tag

Intro + 65% stat + beta availability on Team/Enterprise plans; positions as multiplayer Claude for Slack

02:37 – 03:28

02 · How It Works: Tools & Connectors

Claude uses pre-configured integrations (Gmail, HubSpot, Airtable); responds in Slack threads

03:28 – 04:11

03 · Connect Any App With Zapier MCP

Zapier MCP as bridge to 9,000+ apps for anything not natively connected to Claude

04:11 – 06:26

04 · Claude Is Now Multiplayer

One Claude per channel; shared context; picks up where last person left off; shown via Anthropic diagram

06:26 – 07:38

05 · Ambient Mode: Claude Takes Initiative

Claude proactively flags relevant info across channels and connected tools without being asked

07:38 – 08:29

06 · Ambient Mode Example

Customer login email in Gmail routes to #support and #engineering simultaneously via Claude

08:29 – 09:11

07 · Delegating Work To Many Claudes

Async execution; schedule tasks over hours/days; run many Claudes in parallel

09:11 – 10:40

08 · Privacy & Permissions

Channel-scoped memory; admin controls tools/data per channel; no cross-department bleed; token spend caps

10:40 – 11:42

09 · API Keys, Skills & Plugins

Per-user API keys; Skills/plugins available inside Tag; admin configures which are accessible

11:42 – 13:05

10 · How To Set It Up In Slack

Live walkthrough: add to workspace, grant permissions, @Claude in channel, Slackbot confirmation

13:05 – 14:01

11 · Live Demo: Pulling Emails

Gmail connector demo: @Claude pulls two unreplied morning emails directly inside Slack

14:01 – 15:30

12 · Real Use Case Examples

Anthropic examples: BigQuery spend report, calendar meeting prep, Datadog triage auto-ping

Atomic Insights

Lines worth screenshotting.

Anthropic's own product team generates 65% of its code through Claude Tag — the people who built Claude use it most.
Claude Tag gives each Slack channel one shared Claude instance: everyone picks up the conversation where the last person left off.
Channel-scoped memory means your sales Claude and engineering Claude never share context — by design, not accident.
Ambient mode lets Claude flag a pattern of customer login errors in Gmail and route alerts to both #support and #engineering before anyone notices.
You can connect 9,000+ apps to Claude Tag through the Zapier MCP without writing any code.
Admins can set token-spend caps per user and per channel — critical before you hand a shared AI budget to a whole department.
The Slack launch is explicitly a foundation: Anthropic plans to expand @Claude to Gmail, Microsoft Teams, CRMs, and support ticket software.
Claude Tag is async by design — set a task and focus on other work while it executes, even across hours or days.
Skills and plugins you have configured in Claude carry into Tag — your existing automation stack works inside Slack.
Each channel gets a separate Claude identity: memories, tools, and permissions are scoped to the channels admins define.
The prior Slack Claude bot was a one-on-one chat replacement; Claude Tag is an agentic teammate that the whole channel shares and builds context with over time.
Proactive AI that acts without a prompt is the actual product shift — the Slack surface is just where it lives first.

Takeaway

What changes when your AI is always in the room.

WHAT TO LEARN

Ambient mode is the real unlock — when Claude monitors your channels and acts without being asked, the relationship shifts from tool you invoke to teammate you work alongside.

Shared channel context means Claude accumulates working knowledge over time, so team members do not need to re-explain the project at the start of every interaction.
Ambient mode can surface problems you did not know to look for — a pattern in support emails, a flagged metric in a connected tool — before anyone on the team noticed.
Channel-scoped memory is the privacy mechanism that makes this workable in real orgs: sales, engineering, and support each get a separate Claude identity with no cross-contamination.
Token spend caps per user and per channel are essential to configure before rolling out to a team — without them, one heavy user can exhaust shared budget.
The Zapier MCP bridge means any app Zapier integrates with becomes a Claude data source, extending the practical reach of Claude Tag well beyond Anthropic's native connector list.
Async task execution is already available, but the presenter flags healthy skepticism about how reliable hours-long autonomous runs actually are in practice.
The Slack launch is explicitly positioned by Anthropic as a foundation for expanding @Claude to Gmail, Teams, CRMs, and support platforms — the architecture matters more than the current surface.

Glossary

Terms worth knowing.

Claude Tag: Anthropic's Slack-native AI integration that lets any team member @mention Claude inside a channel, with shared context across the whole channel rather than per-user sessions.
Ambient mode: A Claude Tag feature where Claude monitors connected channels and tools and proactively posts updates or alerts without being explicitly prompted.
MCP (Model Context Protocol): An open protocol that lets AI models like Claude connect to external apps and data sources. The Zapier MCP bridges Claude to 9,000+ third-party services.
Channel-scoped memory: Claude Tag's privacy architecture: each Slack channel gets its own Claude identity with isolated memory and permissions, preventing context from bleeding across departments.
Token spend cap: An admin-set limit on how many API tokens an individual user or channel can consume in Claude Tag, used to control cost in team deployments.
Skills / Plugins: Pre-built instruction sets and resource folders that customize how Claude operates. In Claude Tag, admins configure which skills are available per channel.

Resources

Things they pointed at.

01:03linkClaude Tag Release Article ↗

03:28toolZapier MCP ↗

05:55toolHubSpot CRM

05:55toolAirtable

14:16toolBigQuery

14:26toolDatadog

Quotables

Lines you could clip.

01:00

“The people building Claude in Claude Code are using this 65% of the time now because they've had this for the past couple of months.”

The 65% stat from Anthropic's own team is the most compelling proof point in the video — stops the scroll.→ TikTok hook↗ Tweet quote

04:11

“Claude is now multiplayer. Within a given Slack channel, there's one Claude that interacts with everybody in that channel.”

Clean one-sentence product definition that works as a standalone reel opener.→ IG reel cold open↗ Tweet quote

08:57

“We now spend much of our time delegating tasks to many Claudes in parallel, which over time, this is gonna be more and more how we work.”

Forward-looking claim about the nature of AI-assisted work — punchy and quotable.→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

analogy

00:00Anthropic just released something called Claude Tag, and this is basically a way to use Claude directly inside of Slack with your entire team. Think of this like a multiplayer way to interact with Claude inside of Slack. Anthropic even went ahead and said that the product team is using this 65% of the time when they're using Claude across their entire department.

00:19So in this video, I'm gonna break down everything new in this release, talk about my key takeaways from their release article, as well as I'm gonna show it to you live in action so you could see how it works, and then we're gonna talk about who has access to this starting today. So without further ado, let's dive right into this.

00:34They actually released this article here like they do whenever they release a new product, but I went ahead and broke it down for you guys in a bite sized way so you could see what our main takeaways are from this. Alright. So let's dive right into this.

00:46Anthropic went ahead and said we see Claude Tag as the beginning of an evolution of Claude code and makes the model even more proactive and works better with a full team. Tagging at Claude is now one of the main ways we get things done at Anthropic. Today, 65% of our product team's code is created by our internal version of Claude Tag, which is pretty crazy if you guys think about it.

01:08The people building Claude in Claude Code are using this 65% of the time now because they've had this for the past couple of months. So think of this as, like, Claude code that's built for your entire team that you could trigger directly inside of Slack without you needing to go to the Claude desktop app or needing to go to an IDE like cursor or even the terminal to use Claude code.

01:29Then they go on to say that we're launching Claude Tag on Slack since it's a natural home for collaborative work between teams and AI, and it's where most of Anthropics' day to day work already happens. Right now, it's available in beta for Cloud enterprise and team customers, so you do have to be on either the teams or the enterprise plan in order to access this, which I went ahead and paid for it so I can show it to you guys if you aren't enterprise or team users.

01:52Then they go on to say our goal is to expand where it's available more widely so that teams can tag at Claude in the many other places they work. And this is one of the main points that I wanna, like, hammer home from this release. This is kind of laying the foundation for us to use Claude basically anywhere that we're working, whether that is Slack, Gmail, our specific CRMs that we're using.

02:14Basically, all these different places where we interact with our team, they're planning to roll out at Claude, or I guess you'd call it the Claude tag feature that they're calling it. So, again, as of right now, it's only live on Slack, but imagine this when you have this on, you know, Microsoft Teams or, uh, Claude codes, support your support ticket software, and even on Gmail.

02:33And now we're just using this Claude tag as the central hub where we're able to interact with all these things. Next, they said that if you've worked with Claude code or co work before, Claude tag will feel very familiar. We simply just have to tag at Claude with a request in simple terms, and it'll break down tasks into stages and then work through them in turn using the tools it has access to.

02:56And, again, this is another key point here. Not only are we interacting with this like a chatbot and asking specific questions, and Claude gives us an output, But it's using the tools that we have already preconfigured Claude to connect to, and it's pulling information from those different sources.

03:13Now if you're not already familiar, we could connect Claude to any of the apps that we use on a day to day basis. So whether that's Gmail or HubSpot CRM, Airtable, Spotify, all these different applications so we can now pull data from there directly inside of Slack without us having to go to these apps in particular.

03:29And if there is an app you wanna connect to that you don't see directly inside of Claude, you could always use something like the Zapier MCP, which is something I use every single day, which is basically a bridge between your AI and 9,000 different plus apps that Zapier integrates with. You can simply create an MCP server here even if you're nontechnical, connect the different apps you use like Buffer or Beehive that don't have preexisting connectors inside of the Claude app, and then you just simply click connect, add to Claude, and then now Zapier is connected so we could actually pull from these different apps.

03:59And the more apps that we connect Claude to, the more powerful it's gonna be in order for it to go and get our work done for us. And then they go on to say that once it's done, it'll respond in a Slack thread with what it's created. Alright.

04:12So moving on to kind of the overarching concept of this Claude tag release. Claude is now multiplayer.

04:18Within a given Slack channel, there's one Claude that interacts with everybody in that channel. This means that anyone can see what it's working on, and you pick up the conversation from where the last person left off. This makes Claude very different from working within a single chat or a single task.

04:34It's much more like interacting collaboratively with a teammate. So this feels like this is more of, like, a second brain that your entire team can use. One of the cool things that I've noticed since I've been playing around with this, and I'll show it to you live in action later on in the video, is that you could tell it to, like, summarize an entire thread of things that you missed while you were gone, and it will give you a bite sized breakdown on all of the key things that you missed, things that are, you know, urgent that you need to get back to, a specific message you might have missed.

05:02And for me, when I use Slack personally, things get hidden in all these different threads I didn't even know got started. And now that everybody can use this clog directly inside of your Slack on your team, this can be pretty powerful.

05:12Here is a great visual to break this down a little bit more for you if you're having trouble following along. So Claude is multiplayer, and we we get one Claude per channel that we actually grant access to, and then everybody shares it that's inside of that channel.

05:27So before, we actually had a Claude directly inside of Slack that we could plug in, but this was basically just going back and forth with you and Claude. Instead of you needing to go to the Claude desktop app or claude.ai, you're able just to chat with it right there, but that was not nearly as capable as this new Claude Tag feature.

05:44So now with Claude Tag, we have this Slack channel called product launch with all of our different team members here. We have Sam. We have Sam.

05:52We have Devin. We have Maya. We have Priya.

05:56And now we all have access to this kinda second brain here, um, which is Claude Tag. Alright. So next up, it's saying that as Claude follows along with its channel, it builds more context about the work.

06:06This means that the users don't need to explain things from scratch over and over again. This is pretty cool. I'm excited to see this actually in practice.

06:14And they even went on to mention that Claude can even automatically learn from other Slack channels and data sources only if it's granted permission. Now this right here is probably the coolest feature. Claude is able to take initiative, and they're calling this the ambient mode.

06:29And the TLDR of this is that Claude will proactively keep you updated about whatever it thinks you might need to know. It'll flag relevant information from from across the channels it's in and the tools it's connected to. So when I hear this, I think of something like OpenClaw, where that was kind of the first time we ever saw cron jobs in practice and more of a consumer tool, where it was able to go and say, hey, Brock.

06:51You just got an urgent email. Or hey, I see all of your competitors are making videos about this topic. You should go ahead and do the same.

06:58And since it takes initiative, I don't need to basically send a message to Claude and have it go do a thing. It just does it automatically when it feels it needs to. So, again, here's kind of a diagram, um, visually explaining this.

07:10It's able to watch through these different projects. So imagine we have project x, team sync, and design review as different Slack channels. And not only that, but it also has the tools it's connected to via Claude that I talked about earlier.

07:22Since Claude can always be watching, it will keep you posted on what we need to know. It will flag specific things that matters, and it will be able to actually go and chase loose ends for us. Now a perfect example of this in practice is this for customer support.

07:37So imagine that we give Claude access to our Gmail account, which is used for our customer support. And let's say that there's a couple of emails we get where a customer says they can't log in to our application, for example. Claude will be able to spot this because it has access to this email right here.

07:52Then it will be able to go and alert the right people. And what I mean by alerting the right people is it can go and start a thread inside of the customer support Slack channel in order to flag that with the customer support team, but it can also go ahead and flag this with the engineering team as well so they could go and be proactive and actually try to fix this issue.

08:12And Claude's basically the middleman here, so you don't need somebody who's actively searching in your Gmail to then go and spot this and then alert these different channels inside of Slack. This is just one example that I thought of that could be pretty cool in practice, but I'm really interested to see other use cases of this.

08:28Next, Anthropics said that at Claude works asynchronistically. Set Claude a task, and you could focus on your other priorities while it works. Very similar to, you know, if we've ever used Claude Coworker Claude Code, that is nothing new.

08:40And it says it could even schedule tasks for itself pursuing a project autonomously over hours or days. That sounds obviously pretty promising, but I wanna see this in practice because something tells me that that might not be as capable as they're making it sound right now.

08:55They went on to say that we now spend much of our time delegating tasks to many clods in parallel, which over time, this is gonna be more and more how we work. Instead of us going and doing a thing, it's us delegating that task to our agent, which is now clod tag.

09:10Alright. So let's quickly alright. So now let's talk about privacy and how this actually works, what context does this have, etcetera, because this is something when I originally heard about this release, I thought this was promising, but then privacy instantly became one of the things that I was concerned about.

09:26So what they're saying is that we've designed Claude Tag with teams and organizations in mind. Claude's access to sensitive data and task specific tools can be tightly controlled. To get up and running, system administrators specify which tools and information the model should have access to and which channels.

09:44Think of this like creating separate cloud identities for different uses. Everything including its memories will stay scoped to the channels defined by the administrators. For example, a model set up for sales work won't pass on memories to one setup for engineering.

09:58So whatever department or Slack channel we add this plot tag to, it's only gonna have information and context within that channel. So that way, you don't, like, cross pollinate between your different organizations inside of your company, which you obviously would not want that to be the case.

10:13Once permissions are set, everyone can begin tagging right away, and administrators can set limits for token spend. So this is kind of interesting.

10:22This is something that, obviously, over time, maybe become an issue if there are certain people or certain channels that maybe just spam this for some reason. At least we know that we have the ability to control how much not only organizations and channels are able to use this, but also individual users as well.

10:40Alright. So I dug even deeper into the privacy, and I wanna break this down here because this is pretty important. So Anthropix says, in addition to credentials, admins also define repository access, connectors, which is really big, the tools and API keys that Claude uses to do its job.

10:56And across an organization, different API keys can connect to the same service at different permission levels. I believe this means that each individual user can configure their own API keys for the applications they're using. So that way, the admin doesn't, you know, necessarily have to pay for all of this.

11:12I'm not too sure this is something that I'm not fully in the know of, and I don't wanna speak, you know, incorrectly here.

11:19And skills and plugins, this is really big. So, apparently, we can run our specific skills that we have in plugins. But then the question becomes, does the admin supply the skills and you could only use those, or could each individual use the skills that 've crafted?

11:33I'd probably assume that the admin has to configure it, and it's not just gonna allow all individual users to run their skills, you know, across this Slack channel. But I could be wrong. Alright.

11:42So let's go ahead and set this up now. First of all, there will be a link to this in the description to access this page here. And, again, you do have to be on either the team or the enterprise plan for Claude.

11:52So I'm gonna click add to Slack. And from here, we need to configure which workspace we wanted to have access to. So I'm gonna click AI accelerator, and then we could also manage the permissions here.

12:02So go through and make sure you're only granting access to, you know, certain permissions that you actually wanted to have. I'm just gonna go ahead and click allow. It says Claude is installed, so now I need to open up Slack.

12:12Now what we need to do is come over to admin inside of Slack. So click this little gear icon. And then from here, we're gonna go down to apps and workflows.

12:21So right here, it looks like I actually have Claude added. And if I come back to Slack, I just basically go to any channel that I wanna add Claude to and just do at Claude.

12:30You could see that it's not in the channel. But if I click this, send that off. Slackbot is gonna ask if I wanna add them.

12:35So let's just click add, and we're gonna go ahead and do that for all of our different channels we wanna have access to. Now here, it gives me a message basically breaking down how it works. Here's how I collect and process context.

12:46When you mention me in a thread, I automatically read read recent messages in that thread to understand the conversation context. I can access and analyze files shared in the conversation when you mention me. I only process messages in threads where I'm explicitly mentioned.

13:00Your conversations are private and accessible. I do not retain or train on your Slack conversations. So very interesting.

13:06Alright. So I just gave it a really simple, you know, message here.

13:10I just said, at Claude, how are you doing? It showed that cool little icon showing that it was working. And from here, it shows that I actually need to select a repository that it could work inside of.

13:20So let me go ahead and select one of my repos. Alright. So I asked Claude, how are you doing?

13:24Can you pull my last emails from this morning I have not replied to? And for this use case, I already had Gmail preconfigured inside of my Claude account, so I was able to pull information from that connector that I had.

13:36And so a win pulled two emails that I need to reply to. This is pretty cool using this in practice. And, of course, this is inside of a new Slack channel because I don't wanna show one of the Slack channels inside of my company because there is some private information there.

13:49But as context but as a Slack channel's context builds up, I could see how this can be pretty powerful since it then has general knowledge on all of that information. Alright. So I personally don't have many use cases I could show right now, again, because I don't wanna show sensitive company information, but these are some examples that Anthropic gives us.

14:06So let me just go through a couple of these example. So you could see right here that Elena used at Claude and basically said what are the top enterprise accounts by spend the last seven and twenty eight days. Claude then went and pulled information from BigQuery, which is the application they're using, And then it actually gave a breakdown here.

14:25As you could see, um, it shows all of this ranked along with, you know, the amount of spend over seven and twenty eight days. And on top of that, it even sent a PNG that it went and generated. That's pretty cool, breaking this down with this cool graph here.

14:38Next, this is a pretty simple one. Camilla just tagged it and said, I'm meeting Acme at two. What do I need to know?

14:44And it went and pulled from her calendar, basically breaking down everything that I needed to know, um, in terms of how to prep for that meeting. Nathan sent off a message saying, watch this channel, triage what comes in, and work through the backlog. Tag me only when something needs my call.

14:58So Claude is using the agent to automatically survey this. Looks like there was, like, a checkout error here that Datadog flagged inside of the Slack channel, and then the Claude agent automatically pinged Nathan, um, because this needs his attention. So you could see how powerful this is when we have this monitoring our specific Slack channels and tagging relevant people.

15:16So there we have it. That's a new Claude tag that was just released by Anthropic. If you guys got some value from this video, make sure to subscribe to this channel.

15:23I cover AI for non techies. With that being said, I don't wanna take any more of your time. I hope you have a good day, and I'll see you in the next one.

15:29Cheers.

The Hook

The bait, then the rug-pull.

Anthropic just dropped a stat that reframes the whole Claude Tag announcement: 65% of their own product team's code is now generated through an internal version of this tool. The people building Claude use it most — and that number is the real hook buried inside a feature launch.

Frameworks