Today, Anthropic announced Claude Tag, an always-on instance of Claude that lives inside Slack channels. The idea is that it becomes another participant in a group Slack channel where it can read the channel context, participate in threads, and interact with other channel participants almost as if it were just another employee in the group. Anthropic says 65% of its product team's code is now created by its internal version of Claude Tag.
Claude Tag is designed to function as a persistent teammate that can be summoned via @Claude or, when ambient mode is enabled, proactively engage with ongoing discussions. This marks a significant shift from traditional AI chatbots that require explicit initiation for every interaction. By embedding an agentic AI directly into the communication flow, Anthropic is aiming to reduce friction and increase productivity for teams that rely heavily on Slack for daily operations.
How Claude Tag Works
Claude Tag operates as a single instance of the AI model that interacts with all human participants in a given Slack channel. Anthropic calls this multiplayer mode. Unlike a one-on-one chatbot conversation, Claude Tag sees the entire context of the channel—every message, file, and thread—and can respond to multiple people while maintaining coherent contextual awareness. It knows what it said to Bob, what it said to Sue, and what Bob said to Sue, allowing it to build on collective knowledge without repetition.
When ambient behavior is enabled, Claude Tag can watch the entire channel context over time. It learns from ongoing activity, gathers details from connected tools (such as project management apps or databases), and can proactively follow up on stalled threads or incomplete tasks. For example, if a team member is assigned a task in a channel and hasn't provided an update, Claude might chime in with a gentle reminder or ask for status. This proactive capability can be a double-edged sword: it enhances accountability but may also feel intrusive if not carefully calibrated.
Agentic Capabilities
Claude Tag is not just a passive observer—it is fully agentic. It can break complex tasks into stages, use available tools (like code repositories, spreadsheets, or APIs) to execute work, and then report back in a thread. This asynchronous operation allows users to assign work and move on while Claude works behind the scenes. For instance, a developer could ask Claude Tag to analyze a bug report, search through code commits, and propose a fix—all without leaving Slack.
Anthropic emphasizes that this version of Claude learns over time, continuously tracking what is entered into Slack. According to the company, “This means that users don’t need to explain things to Claude from scratch over and over again.” Over weeks or months, Claude Tag builds a persistent memory of the team’s context, preferences, and ongoing projects, making it increasingly useful. However, the always-on nature raises questions about token consumption, as each ambient action uses computational resources.
Scoped Permissions for Enterprise Teams
Claude Tag is being released exclusively for Claude Enterprise and Claude Team tiers. Anthropic has designed it with organizational security and governance in mind. Each Slack channel gets its own isolated Claude identity. Administrators can tightly scope what each Claude instance can access—including sensitive data and task-specific tools. A Claude in an engineering channel might have access to code repositories and bug trackers, while a Claude in a legal channel might have access to contracts and compliance documents. Importantly, no information is shared between these isolated identities, ensuring data boundaries are respected.
Administrators can also view a full activity log of every action Claude Tag has taken, including which user requested the task. They can set token spend limits at both the organization and channel levels, preventing runaway costs. This financial control is critical because an always-on AI that constantly reads and processes channel activity consumes tokens at a much higher rate than on-demand queries. Anthropic is also offering an introductory launch credit for eligible organizations to help offset early adoption costs.
Claude Tag replaces the existing Claude in Slack app. Existing users will have 30 days to migrate, with Anthropic providing tools to simplify transition. The company notes that the feature is in beta and will expand over time, potentially beyond Slack to other collaboration platforms.
The Broader Context of AI in the Workplace
Claude Tag arrives at a time when enterprises are increasingly exploring persistent AI agents that can operate alongside human teams. Unlike chatbots that respond only when called upon, agentic AIs like Claude Tag can monitor workflows, detect bottlenecks, and even nudge colleagues. This represents a shift from productivity tools that require human initiation to autonomous assistants that integrate seamlessly into communication channels.
However, the always-on model also raises concerns about surveillance and psychological comfort. Some employees may feel uneasy knowing an AI is constantly reading every message—even if it is not reporting back to management. Anthropic has stressed that Claude Tag does not report from private channels and that its activity logs are only visible to administrators, but the perception of an ever-present digital coworker could affect how openly teams communicate.
The company has also addressed token economics. Each channel’s Claude instance, when ambient, consumes tokens to process every message. While administrators can set spending limits, the total cost could still be significant for large organizations with many active channels. Anthropic is betting that the productivity gains—such as faster code development, automated status updates, and reduced context-switching—will outweigh the compute expenses.
Industry analysts note that Anthropic’s move positions it directly against other enterprise AI assistants, such as Microsoft Copilot in Teams and Google Workspace’s Gemini integrations. By focusing on Slack—a platform with a massive user base in tech and media—Anthropic is targeting a core demographic that already embraces AI tools. If Claude Tag proves successful, it could set a new standard for how AI collaborates within messaging apps, potentially prompting competitors to introduce similar ambient agents.
The agentic nature of Claude Tag also aligns with Anthropic’s broader research into safe, capable AI. The company has published guidelines on how its models should handle multi-turn tasks and autonomous tool use. Claude Tag is a real-world test of these principles, and feedback from early adopters will likely shape future iterations. One key area to watch is hallucination: if Claude Tag confidently acts on incorrect information it gathers from a Slack channel, it could lead to errors that are hard to unwind.
Another consideration is integration depth. An always-on AI is only as useful as the tools it can access. Anthropic has not released a full list of supported integrations, but they are expected to include common SaaS platforms like Jira, GitHub, Salesforce, and Google Drive. The more seamless the integration, the more valuable Claude Tag becomes. Conversely, if setup is complex or if APIs are limited, adoption may stall.
Looking ahead, Anthropic plans to expand Claude Tag beyond Slack, potentially to other messaging apps like Microsoft Teams or even email clients. The company has also hinted at a future where different Claude instances in different channels could coordinate with each other under strict administrative controls, enabling cross-departmental workflows without compromising data isolation. For now, however, each channel’s Claude remains siloed, which is a deliberate design choice to prevent unintended data leakage.
Critics point out that relying on an AI to remember context across days or weeks introduces risks if the model forgets or misinterprets earlier messages. Anthropic mitigates this through persistent memory, but the technology is still evolving. Teams will need to decide if the productivity benefits outweigh the potential for miscommunication. The beta period will be crucial for gathering user feedback on accuracy, reliability, and overall user experience.
Claude Tag is available today for eligible organizations. Anthropic invites teams to try it and share their experiences, as the company iterates on the feature based on real-world usage. Whether an always-on AI coworker becomes a beloved addition or a source of friction will likely depend on how thoughtfully it is implemented and culturally adopted within each workplace.
Source:ZDNET News