1300 166 138 Tell us what you need
News / Claude

Anthropic launches Claude Tag, an AI teammate in Slack

By Monika Tantau • Published 5 months ago
Anthropic launches Claude Tag, an AI teammate in Slack

Key takeaways

• Anthropic launched Claude Tag on 23 June, an AI agent you mention in Slack like a team member
• It runs on Opus 4.8, schedules its own tasks and learns context from the channel over time
• Anthropic says 65 per cent of its product team's code now comes from an internal version
• Available in beta on Claude Team and Enterprise plans, and only in Slack for now

Most Read

Anthropic has released Claude Tag, a way to bring its AI model into Slack as a named participant rather than a tool you open in a separate tab. You mention @Claude in a channel and it responds, takes on tasks and keeps working in the background, much as a colleague would.

The beta went live on 23 June for Claude Team and Enterprise customers, according to Anthropic. It is Slack only for now, and admins decide which tools and data each Claude can reach. For Australian teams already living inside Slack, it changes where AI sits in the working day.

What Claude Tag actually does

Until now, using Claude at work usually meant switching to a separate app, pasting in context and copying the answer back. Claude Tag removes that step. The model joins the channel, reads the thread it is mentioned in and acts on the request without anyone leaving Slack.

Because it sits where the conversation already happens, it can pick up tasks that span several people. Anyone in the channel can see what it is doing and carry the work forward, instead of each person holding a private chat with their own assistant.

One shared agent, not many chatbots

The design choice that stands out is a single Claude per channel, shared by everyone in it. That makes its work visible and collaborative, closer to a junior team member than to a private conversational AI window.

It also behaves agentically. Claude Tag can schedule its own tasks, work asynchronously and build context about the channel over time. It runs on Opus 4.8, and administrators control its identity and access for each channel, so a finance channel and a support channel can be scoped very differently.

The 65 per cent claim

The headline figure in the announcement is Anthropic’s own. It says 65 per cent of its product team’s code is now created by its internal version of Claude Tag. That is a striking number, though it describes a software company full of engineers who build with these tools every day, not a typical mid-market operation.

We would read it as a direction of travel rather than a benchmark to hit. The more useful question for most businesses is narrower: which repetitive, routine jobs in a channel could a shared agent reliably take off people’s plates?

Why Australian businesses should pay attention

Putting an AI agent inside Slack with access to company tools raises real governance questions. The agent can act, schedule and reach data, so the access an admin grants matters as much as the model behind it. Under the Privacy Act 1988, Australian firms stay responsible for how customer data is handled, even when an agent touches it.

This is where deployment beats novelty. Deciding which channels get a Claude, what it can see and where its outputs go is an AI strategy question, not just an IT toggle. Teams that scope it tightly will get the productivity without the exposure.

Why Australian businesses should pay attention

Where agentic AI is heading

Claude Tag is part of a wider shift, from chatbots you visit to agents that sit inside the tools you already use. The major providers are all moving this way, and for anyone weighing the options, our comparison of Claude versus ChatGPT for business workloads is a useful starting point.

For Australian teams, the takeaway is simple. The technology to put a capable agent in your daily workflow now exists. Whether it helps or creates risk depends almost entirely on how carefully you set it up.

Scroll to Top