Claude Tag: Anthropic's Slack AI Coworker, Explained

> **TL;DR:** Claude Tag is a new Anthropic feature that lets teams @-mention Claude directly inside a Slack channel, where it works on tasks in public view, posts progress updates, and can even run unprompted on a schedule. Anthropic says it now uses Claude Tag internally so heavily that 65% of its product team's code is written by it, though the feature itself is closed and paid — which has already prompted a free, self-hosted, model-agnostic alternative to emerge.

Key Takeaways

- Claude Tag lets anyone in a Slack channel @-mention Claude to assign it a task, turning it into a visible shared teammate instead of a private 1:1 chatbot. - It supports an autopilot mode: scheduled status reports and unprompted investigation of issues without a human triggering each run. - Claude Tag accumulates memory of how a team and company operate, refining its usefulness over time rather than starting cold each session. - Anthropic says 65% of its own product team's code is now written through this workflow — a striking internal-adoption data point. - The feature is closed and paid, which has already spurred a free, open-source, self-hosted alternative that works with Claude, GPT, Gemini, or local models.

What Is Claude Tag?

Claude Tag is Anthropic's answer to a problem every team using AI assistants eventually hits: work done in a private chat window is invisible to everyone else. Instead of opening a one-on-one conversation with Claude, Claude Tag lets anyone in a Slack workspace simply @-mention Claude in a channel — the same way you'd tag a colleague — to hand it a task. Claude then works the problem in that channel, posting progress updates as it goes, so the whole team can watch the task move from assigned to done.

That shift from private chatbot to public teammate is the entire pitch. A task delegated to Claude Tag isn't a black box between one employee and a model; it's a thread anyone can follow, question, or redirect in real time.

How Claude Tag Works Inside Slack

The mechanics are deliberately close to how you'd manage a human report. You tag Claude in a channel with a task description, and it starts working — posting updates as it makes progress rather than going silent until a final answer. Because the whole exchange lives in a shared channel, teammates can jump in mid-task with clarifications or new context without breaking the workflow.

Anthropic has also built in an autopilot mode, where Claude Tag doesn't wait to be summoned at all. It can send scheduled status reports on its own cadence, or proactively investigate an issue it notices — no one has to remember to ask. Over time, it builds up memory of how the company actually operates: its tools, its conventions, its recurring problems. That accumulated context is arguably the more interesting part of the feature, since it means Claude Tag should get more useful the longer a team runs it, rather than resetting to zero each session.

Anthropic Is Reportedly Its Own Biggest Customer

Anthropic says it's already using Claude Tag heavily inside its own product team, and the adoption number it's citing is hard to ignore: 65% of that team's code is now reportedly written through this workflow. That's not a pilot-program statistic — it describes a team where the majority of code output is already flowing through an @-mentioned AI teammate rather than through engineers typing every line themselves.

It's a useful data point for any organization trying to gauge how far "AI coworker" tooling has actually progressed, separate from vendor marketing. Anthropic is, in effect, dogfooding Claude Tag at a scale that suggests real workflow dependence, not just novelty usage.

Why This Matters Beyond Anthropic

The number lands at a moment when the AI agent conversation has moved past "can it write a function" and into "can it own a piece of team process." Claude Tag's design — public visibility, autopilot investigation, persistent memory — is aimed squarely at that second question. For more on how agent tooling is evolving across the ecosystem, see our recent roundup of [GitHub's weekly wins in AI agent repos](https://speka.info/blog/github-weekly-wins-13-repos-reshaping-ai-agents).

The Catch: Closed, Paid, and Anthropic-Only

Claude Tag is not an open standard or a free add-on — it's a closed, paid feature tied to Anthropic's own product. That's a reasonable business decision for Anthropic, but it also means any team that wants this exact workflow is committing to a single vendor, a subscription cost, and Anthropic's infrastructure for handling what could be sensitive internal task data and company context.

For smaller teams, indie developers, or companies with strict data-residency requirements, that combination — pay, and hand your operational memory to a third party — is a real barrier, not just a preference.

Enter the Open-Source Alternative

That gap is exactly what a new free, open-source project is targeting. It reproduces the core Claude Tag concept — a shared AI teammate that lives in a chat channel and works on tagged tasks in full view of the team — but as something you self-host rather than subscribe to.

Model-Agnostic by Design

The key structural difference is that the open-source version isn't tied to a single model provider. It's built to work with Claude, GPT, Gemini, or locally-run models, so teams can plug in whichever model fits their budget, latency, or compliance needs — or swap models later without rebuilding the whole workflow. That flexibility matters given how quickly the underlying model landscape keeps shifting; see, for instance, the debate our coverage of [GLM 5.2 and the 'AI margin collapse' discussion on Hacker News](https://speka.info/blog/glm-5-2-sparks-ai-margin-collapse-debate-on-hn) captured about how fast model economics are moving.

Why Self-Hosting Matters for Data Governance

Because it's self-hosted, the project avoids recurring subscription costs and — just as importantly — keeps company data on private servers rather than routing it through an external vendor's systems. For teams whose "Claude Tag equivalent" would be reading internal tickets, code, and status conversations, that data-residency difference is often the deciding factor over which option to adopt, independent of feature parity.

Claude Tag vs. the Open-Source Alternative

Put side by side, the trade-off is straightforward:

- **Anthropic's Claude Tag**: polished, tightly integrated with Claude specifically, backed by Anthropic's own infrastructure and (per its internal usage) proven at real scale — but closed, paid, and single-vendor. - **The open-source alternative**: free and self-hosted, model-agnostic across Claude, GPT, Gemini, or local models, and keeps data in-house — but requires a team to stand up and maintain the hosting themselves.

Neither option is strictly better; they optimize for different constraints. A team that wants a turnkey, vendor-supported experience and is comfortable working exclusively in Claude has a clear reason to pay for Claude Tag. A team that wants model flexibility, tighter data control, or simply doesn't want another subscription line item has an increasingly credible free path to the same workflow.

What This Means for Teams Evaluating AI Coworkers

The broader signal here is that "AI as a tagged, visible teammate in chat" is becoming a category, not a one-off feature. Anthropic's internal 65% figure suggests the pattern can scale to a meaningful share of real engineering output when a team commits to it. And the fact that an open, model-agnostic alternative appeared quickly enough to matter suggests this workflow pattern — not just Claude Tag specifically — is what's catching on.

For teams evaluating either path, the practical questions are the same regardless of which one you pick: how much autopilot autonomy do you want Claude (or whichever model) to have before a human reviews its work, how much internal memory are you comfortable letting it accumulate, and whether your data governance requirements point you toward a hosted vendor or a self-hosted setup.

Where This Fits in the Bigger Agent Trend

Claude Tag arrives alongside a steady drumbeat of Anthropic product news — see also our coverage of [Claude's Fable 5 personality returning globally on July 1](https://speka.info/blog/anthropic-fable-5-returns-globally-july-1) — that together paint a picture of Anthropic pushing Claude further into daily team workflows, not just as a chat assistant but as infrastructure. Teams tracking this space can find more coverage of new releases and workflow tools in our [New AI Tools & Skills hub](https://speka.info/new-ai-tools/).

Frequently Asked Questions

What is Claude Tag?

Claude Tag is an Anthropic feature that lets Slack users @-mention Claude in a channel to assign it a task. Claude then works publicly in that channel, posting progress updates, rather than in a private one-on-one chat.

Can Claude Tag work without being asked each time?

Yes. Claude Tag has an autopilot mode where it can send scheduled status reports or investigate issues on its own, without a human prompting each run.

Is Claude Tag free?

No. Claude Tag is described as a closed, paid Anthropic feature. A free, open-source, self-hosted alternative with similar functionality has since emerged.

How much of Anthropic's own code is written using Claude Tag?

Anthropic reportedly uses Claude Tag internally to the point that 65% of its product team's code is now written through this workflow.

What's different about the open-source alternative to Claude Tag?

It's self-hosted and model-agnostic, working with Claude, GPT, Gemini, or local models, and it avoids subscription costs by keeping data on private servers instead of a vendor's infrastructure.