AgentTeams vs Claude Tag: One Teammate or a Whole Team?

On June 23, 2026, Anthropic launched Claude Tag: a persistent AI teammate that lives inside your Slack channels. You tag @Claude into a channel, it picks up the shared context around it, and it takes on real work using the tools your organization has connected. It remembers what happens in its channels, follows up on its own, and replaces the old Claude in Slack app. It is a genuinely strong release, and it is the clearest signal yet that AI teammates, not AI chatbots, are the direction the whole category is heading.

So the question we started getting within hours: how does AgentTeams compare? The honest answer is that they are different shapes of the same idea. Claude Tag gives you one assistant that lives in Slack. AgentTeams gives you a team of AI employees that works across every tool you use, including the ones facing your customers. Both are good. Which one fits depends on where your work actually happens and how much of it you want handled.

What Anthropic launched

Claude Tag turns Claude into a standing member of a Slack channel. Grant it access to selected channels and connect it to your tools, and anyone in the channel can tag it in and hand off a task. It builds context as it follows along, so people do not have to re-explain the work each time. With its ambient mode switched on, it proactively flags things it thinks you need to know and chases threads that have gone quiet. It can even schedule tasks for itself and work a project over hours or days.

The administrative controls are thoughtful. Admins decide which tools, data, and channels each instance can touch, set spend limits, and review a log of everything it did and who asked. Memory stays scoped to its channels, and separate instances stay walled off from each other, so a setup for sales will not leak memories or data into a setup for engineering. It is in beta for Anthropic Enterprise and Team customers, billed by usage. If your company already runs on Anthropic and lives in Slack, this is an easy, powerful addition.

One teammate versus a team

The first real difference is headcount. Claude Tag is one assistant per channel. Within a channel there is a single Claude that everyone shares and talks to. That is a clean model for a working group that wants one capable helper on call.

AgentTeams is built around the opposite idea: a team of distinct AI employees, each with a role, a personality, and a job to do. You have a support agent, a sales agent, an operations agent, and they hand work to each other the way colleagues do. A support agent that hits a billing question can pass it to the finance agent and loop back with the answer. That is the premise behind multi-agent teams: specialized roles that coordinate beat one generalist trying to be everything to everyone.

Slack versus everywhere your work happens

The second difference is reach. Claude Tag lives in Slack. That is exactly right for internal collaboration, where Slack is the office. But a lot of real work does not happen in Slack at all. Your customers email you, open support tickets, and message you on chat. Your team might run on Telegram or Google Chat instead. An assistant that only lives in Slack cannot see or touch any of that.

AgentTeams agents work across every channel at once, and remember the same person from one to the next. The same agent can answer a customer in your support inbox, pick the conversation back up over email, and report to your team in Slack, without losing the thread. This is the cross-tool model we made the case for in horizontal vs vertical AI agents: the value compounds when one agent works across your tools and remembers across them, rather than being trapped in a single app.

Tag a bot versus hire an employee

The third difference is the mental model. Claude Tag is something you tag in, an assistant you call on inside a conversation. AgentTeams is something you hire. You give an agent a role, brief it with directives the way you would brief a new employee, connect its accounts, and set approval gates so drafts wait for your sign-off until you trust it to run on its own. You manage it, review its work, and grow its responsibilities over time. That is the whole idea of an AI employee platform: the people running it are managers, not engineers.

Subscription versus a usage meter

The fourth difference is the bill. Claude Tag is billed by usage. Channel work is charged to the organization and direct messages to the individual, so every conversation your teammate has adds to a running meter. For light or occasional use that is perfectly reasonable. For an assistant that is busy all day, the meter never stops, and the cost is hard to predict from one month to the next.

AgentTeams can run on a subscription you already pay for, not a usage meter. Point your agents at the Claude or ChatGPT plan your team already has, and they work against usage that plan already includes, so your cost is a flat, predictable line instead of a tally of every message. For teams running agents every day, that is usually far less than pay-as-you-go, and far easier for finance to plan around. We broke the math down in running agents on your own subscription.

Where Claude Tag clearly wins

If your company is already an Anthropic Enterprise or Team customer and your work lives in Slack, Claude Tag is a very easy yes. It is first-party, so you get the latest model with no setup, no separate vendor, and billing that lands on a bill you already pay. For internal engineering and product work in Slack, it is hard to beat. Anthropic uses it heavily in-house, and a strong assistant that is one tag away from every channel is a real productivity unlock for teams that already work that way. If that describes you, start there.

Where AgentTeams clearly wins

AgentTeams wins when the work is bigger than one assistant in one app. You want a whole team of specialized agents, not a single generalist. You want a flat, predictable cost on a subscription you already pay for, not a meter that ticks with every message. Your work spans more than Slack: Slack and email and support tickets and chat, often facing customers, not just teammates. You want to manage agents like employees with roles, directives, and approval gates rather than tag a bot per conversation. And you do not want to be locked to one model provider or one chat app to get there. For customer-facing support, sales, and operations that cross several tools, a team you hire beats an assistant you tag.

The bottom line

Claude Tag is the right call for internal teams that live in Slack and want one excellent assistant on tap. It validates everything we believe about where this is going: AI that remembers, takes initiative, and works on its own, under real controls.

AgentTeams is the right call when you want an actual workforce. A roster of specialized AI employees, working across every tool and channel your business runs on, handling customer-facing work as comfortably as internal work, managed the way you would manage people. If your question is "how do I get one great assistant in our Slack," tag Claude in. If your question is "how do I get a team handling support, sales, and operations across all our tools," you want a team you can hire.