Anthropic’s Claude Tag Plants a Persistent AI Teammate in Slack

Anthropic launched Claude Tag on June 23, the company’s first product built around an “always-on” version of its Claude model that lives inside Slack channels as a shared teammate. Available in beta for Claude Enterprise and Claude Team customers, Claude Tag replaces Anthropic’s existing Claude in Slack app and runs on Claude Opus 4.8, the model Anthropic put out less than a month ago.

The visible story is a new product in a crowded market for AI assistants inside team chat. The consequential story sits underneath it. Anthropic is pairing Claude Tag with a new access architecture it calls “agent identity,” which moves AI permissions from per-user to per-channel and turns the channel itself into a repository of vendor-specific institutional memory. That has implications well beyond a product launch.

What Claude Tag Actually Does

Inside a Slack workspace, an administrator pairs Claude Tag with the workspace, grants it access to specific tools and data sources, sets a token-spend ceiling, and defines which channels it can operate in. From that point on, any team member in those channels can tag @Claude with a request to write a pull request, pull a sales number, or run an analysis, and Claude will break the task into stages, execute them through the tools it has been granted, and post the result in the Slack thread. The four-step setup is designed to minimize friction for IT teams already managing sprawling SaaS portfolios.

Four behaviors distinguish Claude Tag from the integrations that preceded it. It is multiplayer: within a given Slack channel, there is one Claude that interacts with everyone, and any team member can see what the bot is working on and pick up the conversation from where the last person left off. It learns over time: as Claude follows its channel, it accumulates context about the work and stops asking users to re-explain projects, and it can also draw on information from other Slack channels if it has been granted permission. It doesn’t, however, report from private channels. It takes initiative when “ambient” behavior is turned on, surfacing information it thinks the team needs and following up on threads that have gone quiet. And it works asynchronously, with Claude able to schedule its own tasks and run them over hours or days.

The same pattern is already running inside Anthropic’s own walls. Today, 65% of the company’s product team code is created by its internal version of Claude Tag, and the same setup is being used to chase product metrics, work through support tickets, and root-cause production bugs. Existing customers have 30 days from the June 23 launch to migrate from the prior Claude in Slack app, and Anthropic is issuing introductory launch credits to eligible Enterprise and Team organizations.

The Agent Identity Model Underneath

Claude Tag is built on a different access model from the standard “act as the user” pattern that most AI assistants still follow. Anthropic calls the new model agent identity, and it gives Claude its own accounts for each system it touches rather than borrowing the credentials of whichever human invoked it.

In practice, that means Claude posts in Slack as the Claude app, opens pull requests under its own GitHub App, and queries a data warehouse through a service account that an admin provisions. Memory and tool access are scoped per channel, with admins defining a baseline identity at the workspace level and then tightening or loosening access for individual channels. Anthropic’s own example is the cleanest: an agent identity configured for sales work cannot pass memories to one configured for engineering, and an engineer cannot ask a sales-channel Claude to reach into a private legal document.

The reason for the new model is twofold. First, agents have become more autonomous: the length of a task an AI agent can reliably complete on its own has been doubling roughly every four months, and Claude now schedules tasks for itself and acts on events long after the person who asked has logged off. Second, the channels Claude Tag lives in are multiplayer, with three engineers and a product manager steering the same bot at the same time. When more than one person is steering, there is no single user whose credentials make sense to inherit. The old “act as the user” pattern breaks down, and per-channel scoping takes its place.

The shift from single player to multiplayer AI in products like Claude Tag makes long-running, team-based work possible. Agent identity ensures that Claude’s access to tools is broad enough to be useful, but scoped enough to be secure at enterprise scale.

That is from Noah Zweben, a member of technical staff on Anthropic’s Claude Code team, writing in Anthropic’s agent identity access model.

Why Anthropic Is Building a Permanent Seat at the Channel

The financial stakes behind a Slack-resident AI are large enough to explain the build-out. Anthropic raised $65 billion in Series H funding in late May at a $965 billion post-money valuation. The valuation landed one week before the Claude Tag launch, in Anthropic’s confidential IPO filing. Claude Code’s run-rate revenue alone has grown to over $2.5 billion, more than doubling since the start of 2026, with enterprise use accounting for more than half of that total. The economics matter because a Claude with persistent access to a Slack channel, connected tools, and ambient monitoring represents a much deeper integration than a chatbot conversation or an API call. Token consumption rises, usage patterns get stickier, and the switching costs accumulate.

Enterprise scale is already arriving. Anthropic’s largest enterprise AI deployment to date is Deloitte’s rollout of Claude to more than 470,000 employees across 150 countries, a deployment that crossed the threshold last year and has been Anthropic’s most-cited reference customer since. Claude Tag is positioned as the next layer on top of that footprint, with the company explicitly framing it as the beginning of an evolution of Claude Code rather than a side product.

• $965 billion: Anthropic’s post-money valuation from its late-May Series H.
• $47 billion: Anthropic’s run-rate revenue as of June 2026.
• $2.5 billion: Claude Code run-rate revenue, more than double its start-of-year figure.
• 470,000+: employees at Deloitte with access to Claude under the largest enterprise rollout to date.
• 40%: share of enterprise applications Gartner forecasts will feature task-specific AI agents by 2026, up from less than 5% in 2025.

The wider market numbers reinforce the bet. Fortune Business Insights projects the global agentic AI market will grow from $9.14 billion in 2026 to $139 billion by 2034, and Anthropic’s Claude Tag launch announcement positions the product as the company’s most aggressive move yet to colonize the enterprise collaboration layer.

The Race for the Slack Layer

Claude Tag lands in a market that has tightened sharply over the past six months. Salesforce, which acquired Slack for $27.7 billion in 2021, announced more than 30 new capabilities for Slackbot in March, the most sweeping overhaul of the platform since the deal. OpenAI introduced “Workspace Agents” in April, letting enterprise subscribers design agents that take on tasks across Slack, Google Drive, Microsoft apps, Salesforce, and Notion. Perplexity launched an enterprise “Computer” agent with direct Slack integration. Cognition’s Devin has been built around Slack as a primary interface since its early days, and Microsoft has brought GitHub Copilot into Teams.

The strategic logic is straightforward. The average enterprise juggles over 1,000 applications, and employees waste hours on context switching between them. Whichever AI system becomes the default presence in the channel where work is coordinated gains an enormous distribution advantage and, critically, an enormous data advantage, because the AI that lives in the channel absorbs the institutional context that becomes hard to replace.

Anthropic isn’t the only company chasing the same layer. Microsoft’s bet runs through Copilot and Microsoft’s Work IQ intelligence layer, the foundation for agents that carry chat, context, tools, and workspaces through A2A, MCP, and REST APIs. Snowflake and Databricks are positioning their data platforms as back-end support for the tacit organizational knowledge agents tap into, and Glean is building an intelligence layer that sits between the model and enterprise data. The pattern is the same across all of them: the vendor that owns the agent in the channel owns the context, and the context is what makes the agent worth keeping.

The implication for enterprise procurement is direct. Picking the agent that lives in the chat channel is a multi-year commitment, and the conversation about which AI vendor to standardize on is now a conversation about which AI vendor is already inside the channel.

The Risks Inside a Permanent AI Colleague

The first risk is vendor dependency. Once an organization’s agents, operational configurations, and channel context run on Anthropic’s managed infrastructure through Claude Tag, switching costs rise sharply. A Claude that has accumulated months of channel context and institutional memory becomes very difficult to replace, and Anthropic is explicit in its own product blog that the value of the tool “compounds with tool and context access.” Enterprise procurement teams accustomed to multi-cloud flexibility will need to think hard about what it means to hand a single vendor’s AI persistent access to the communication layer where institutional knowledge lives.

The second risk is governance around ambient monitoring. Claude Tag’s ambient mode, in which the bot proactively decides what its human teammates need to know, is a meaningful expansion of what enterprise AI systems do. For regulated industries, an AI that monitors information flows and makes editorial judgments about what humans should see raises questions that existing AI governance frameworks have not yet addressed.

The third risk is pricing. Anthropic hasn’t published detailed pricing for Claude Tag beyond noting that it runs on token-based spending with administrative controls, and the token-consumption profile for an agent monitoring channels continuously and working asynchronously over hours or days could look very different from a chatbot session. Anthropic says it plans to expand Claude Tag beyond Slack once the beta settles, which means the lock-in question will travel with the product into whatever channels come next.

Frequently Asked Questions

What is Claude Tag?

Claude Tag is an Anthropic service that embeds Claude directly inside Slack channels as a persistent shared teammate. Any team member can tag @Claude with a task and watch Claude work through it inside the channel using the tools it has been granted.

How is Claude Tag different from the existing Claude in Slack app?

The original Claude in Slack app let users DM the bot or invoke it on demand in a thread. Claude Tag replaces it with a single Claude identity per channel that builds memory over time, runs autonomously, and can be steered by any team member who walks into the conversation.

When and for whom is Claude Tag available?

Claude Tag entered beta on June 23, 2026, for Claude Enterprise and Claude Team customers. Administrators have 30 days from launch to opt in to migration from the older Claude in Slack app.

What controls do administrators have?

Administrators define separate Claude identities per use case, set channel-by-channel boundaries on tools and data, set organization-wide and per-channel token spend limits, and view a complete audit log of every action Claude takes and which user requested it.

What model does Claude Tag run on?

Claude Tag runs on Claude Opus 4.8, the model Anthropic released in late May 2026 with improved agentic coding and knowledge-work benchmarks.