Claude Tag launched on June 23, 2026. It is Anthropic’s attempt to build an AI agent that lives inside Slack — one that works in channels, remembers context over time, and handles tasks when you tag it.

I use Viktor. I use it every day. I have built a significant part of my publishing operation around it. So when Anthropic announced a direct competitor, I paid attention.

What I found is instructive — not just about Claude Tag, but about what it actually takes to build something that works.

Claude Tag is a new feature from Anthropic that allows users to call @Claude into any Slack channel. Once invited, it can read conversations, take on delegated tasks, remember what it has learned about your organization, and follow up without being prompted again.

Anthropic says 65% of its own product team’s code is now generated by an internal version of Claude Tag. The company is positioning it as a genuine team member — persistent, context-aware, proactive. It is in beta for Claude Enterprise and Claude Team customers. The existing Claude in Slack app will be switched off entirely on August 3, 2026.

Those are the facts. The experience, it turns out, is a different matter.

Independent testing began almost immediately after the launch. The verdict, from multiple users across different teams, was consistent.

“I’ve been testing Claude Tag today,” wrote one user shortly after launch. “Its not got s**t on Viktor. Its quite meh. The features are all there, but it just feels subpar on quality and delivery.”

That summary deserves unpacking, because it captures something precise. The features are all there. But quality and delivery are where it falls short. That gap is not an accident.

The same tester went further. Claude Tag felt chaotic and messy, and lacked the clear orchestration Viktor brings. He had to explicitly ask it to be proactive — by default, it reverted to behaving like a chatbot, waiting to be asked rather than working ahead. He struggled to get it to engage with Slack channels properly. And it could only be addressed from within the specific channel it was assigned to — attempting to coordinate across channels was blocked.

“It’s also very chatty when it doesn’t need to be,” he added.

This is a distinction that matters enormously in AI tools, and it is one the industry consistently underestimates.

A feature list is straightforward to build. A list of capabilities — channel awareness, memory, task delegation, asynchronous follow-up — can be assembled and shipped. Anthropic has shipped exactly that. The announcement is technically accurate.

Execution is different. Execution is what happens when a real person, with a real workflow, tries to use those features under real conditions. And real conditions include: not remembering to prompt the agent to be proactive, needing to coordinate work across multiple channels, expecting the tool to show judgment rather than wait for instruction.

Viktor passes that test. Claude Tag, in its current form, does not.

The word that appeared in independent testing was “orchestration.” Viktor brings clear orchestration. Claude Tag lacks it.

Orchestration in an AI agent means the tool understands the structure of work — not just individual tasks, but the sequence, priority, and dependencies between them. It means the agent makes sensible decisions about when to act, when to report back, and when to wait. It means the experience is coherent rather than reactive.

Viktor was built for this from the start. The architecture is designed around autonomous work — skills that persist, integrations that connect across tools, a Slack presence that understands it is a coworker in a workspace rather than a chatbot inside a single thread. Viktor launched in February 2026, went viral with 2.5 million views in its first week, and was approved by Salesforce’s security team for the Slack Marketplace in two weeks — a process that typically takes two to three months.

That approval matters. Of approximately 2,900 apps in the Slack Marketplace, fewer than 50 are AI agents. Viktor is the only general-purpose AI coworker with over 3,200 integrations. Claude Tag, at present, is enterprise-only, beta-only, and works inside Slack but cannot connect out to that breadth of tools.

One specific finding from testing is worth addressing directly: Claude Tag defaults to chatbot behavior.

A chatbot waits. It answers questions when asked. It processes one request, delivers one response, then stops. That is the standard mode for almost every AI tool released in the last three years — and it is the mode most users encounter when they open a new AI product for the first time.

An agent is different. An agent holds context, identifies what happens next, and moves the work forward without being asked. The difference in practice is significant. With a chatbot, you do the thinking. With an agent, you share the thinking.

Claude Tag, according to the testing, defaults to chatbot mode. It has agent capabilities — but they require explicit activation. Users had to ask it to be proactive. That is, in effect, asking the agent to be an agent. Viktor does not require that instruction. Autonomous behavior is the starting point, not an option you have to unlock.

The arrival of Claude Tag is, in one sense, confirmation. Anthropic — a company with billions in funding, a flagship language model, and deep integration into enterprise technology — has concluded that the future of AI at work is an agent in Slack. That is the same conclusion Viktor reached, and built for, at launch.

The question for anyone evaluating tools right now is not whether to use an AI agent. That decision has effectively been made by the market. The question is which one actually delivers when the task is real and the day is busy.

Anthropic has the features. It does not yet have the execution. The gap — the one that matters to a real user running a real workflow — is in orchestration, proactivity, and coherence. Those are not small differences. They are the difference between a tool you use and a tool you forget you have.

Viktor is the AI agent I use and recommend. Viktor runs on the three most powerful AI models in the world — Claude, GPT-4, and Gemini — and selects the right one for each task automatically. You do not need to know which is which. You give Viktor an instruction in plain English, and it delivers.

You get $100 of free credits to begin — no credit card, no time limit, no commitment. Explore Viktor properly. Do real work. When you are ready to go further, $50 comes straight off your first bill.

Start here: https://ref.viktor.com/alun-hill

The lesson here is not that Anthropic has failed. Claude Tag will improve. Anthropic has the resources and the model quality to iterate quickly.

The lesson is that in software tools, feature parity and product quality are not the same thing. Getting all the features working is the engineering problem. Getting them to feel coherent, intentional, and trustworthy in daily use is the product problem. The product problem is harder.

Viktor solved it first. That head start is not easily closed.

Disclosure: Some links in this article are affiliate links. If you choose to get started with Viktor using the links provided, I may receive a commission — at no additional cost to you. I only recommend tools I use and believe in.