Key Points:
- Anthropic released Claude Tag in beta, embedding a persistent, always-on artificial intelligence teammate directly inside Salesforce’s Slack.
- Unlike traditional chatbots, the agent reads channel context, participates in group threads, and acts as a shared coworker for entire teams.
- An optional “ambient behavior” mode allows the AI to proactively surface relevant cross-channel updates and follow up on stalled tasks.
- Administrators maintain tight security controls, establishing strict data boundaries to prevent sensitive information leaks between departments.
Artificial intelligence pioneer Anthropic PBC has officially launched a highly anticipated new product that embeds its advanced models directly inside the corporate collaboration layer. Dubbed “Claude Tag,” the newly released tool is currently in research preview for Salesforce’s Slack platform. Unlike traditional chat assistants that operate on an isolated, one-on-one basis, the new agent integrates directly into shared communication channels, functioning as an always-on, persistent virtual teammate. By simply typing @Claude in a Slack thread, any member of a corporate team can bring the AI into group chats, fundamentally shifting how businesses manage project context, delegate tasks, and organize institutional knowledge.
For enterprise technology leaders who have spent the past two years determining how to leverage machine learning within their daily operations, the release of this collaborative agent reframes the conversation entirely. This is not a simple question-and-answer chatbot, an isolated coding assistant, or a basic search tool bolted onto a messaging app. Instead, Claude Tag represents a major milestone in the evolution of agentic AI. It is designed to function as a standing, collaborative member of a department—one that accumulates memory, takes initiative, works asynchronously, and interacts publicly with every person in a channel rather than serving a single user.
At its core, the new platform introduces two major technical breakthroughs designed to revolutionize collaborative workflows. The first is true multi-user collaboration: within designated communication channels, a single, shared instance of the AI interacts with all team members. Because the agent publishes its work product, code segments, and research logs directly into the shared channel thread, everyone can monitor its real-time progress. This public execution allows team members to seamlessly continue conversations where others left off, transforming the AI from an isolated personal assistant into a highly synchronized, shared colleague.
The second breakthrough is continuous context accumulation. Traditional artificial intelligence tools suffer from short-term memory limitations, requiring users to repeatedly copy-paste background documents, upload files, and write tedious introductory explanations before assigning a new task. In contrast, Claude Tag builds context dynamically by constantly reading and following along with the active conversations in the channels where it resides. As it observes daily team workflows, the agent naturally learns more about the specific metrics, terminology, and operational standards of the department, drastically reducing the need for repetitive onboarding explanations over time.
To further elevate team productivity, the agent features an optional “ambient behavior” mode that allows the AI to engage proactively without being explicitly prompted. When administrators enable this feature, the system goes beyond merely responding to user requests, actively monitoring broader organizational updates. It can automatically surface relevant facts from other public channels, flag forgotten projects that have not received an update in several days, and jump into active threads to offer timely data insights. This proactive engagement turns the AI into a highly watchful supervisor, ensuring that critical team tasks do not quietly slip through the cracks.
Because granting an AI agent the ability to read corporate channels and access internal databases raises immediate security concerns, Anthropic has built robust governance controls directly into the platform’s architecture. Systems administrators retain tight, granular control, specifying exactly which channels, tools, and internal databases the agent can access on a per-channel basis. Each instance of the AI maintains completely isolated memories and permissions. This strict division ensures that a specialized instance configured for legal compliance cannot share private memories with an engineering channel, completely preventing the cross-contamination of sensitive corporate data.
Under the hood, the new collaborative agent is powered by the company’s most advanced, recently updated foundation model, Opus 4.8. This high-capacity model enables the agent to break down complex, multi-stage assignments into manageable milestones and execute them autonomously over several days. The tool officially replaces the older, highly restricted “Claude in Slack” application that was first introduced last October. By launching the tool in a research preview, the company is targeting subscribers of its premium Claude Enterprise and Claude Team plans, which are designed to handle the rigorous security and high-volume demand of global corporations.
The company is backing the commercial release with striking real-world data gathered from its own internal operations. According to official release notes, an astounding 65% of all code currently merged into Anthropic’s own product codebase is authored and verified using internal versions of the collaborative agent. The startup also relies on the tool to run its internal support systems, analyze real-time product metrics, and identify complex software bugs. This massive internal adoption serves as a powerful proof of concept, demonstrating that the very engineers who built the model are already relying on it to handle the majority of their daily coding and development tasks.
As the company rolls out the research preview to its enterprise subscribers, this launch is being viewed as a critical step in a broader corporate expansion that has driven Anthropic’s market valuation to a staggering $965 billion. By integrating its agent directly into the communication layers where corporate decisions are made, the startup is positioning itself to dominate the next era of enterprise software. If the collaborative agent can successfully prove its worth in everyday corporate environments, it will pave the way for a permanent reorganization of the modern workplace, where humans and secure, autonomous AI agents work alongside each other as equal teammates.
