Anthropic Launches Claude Tag: AI Lives in Slack

Claude Tag layers persistent context and memory on top of that, something the previous tools could not easily maintain. In Anthropics words, as Claude follows along with its channel, it learns ever more about the work. It can also automatically gather facts from elsewhere in the organization, if it is granted permission to read other channels.

Three things make this version feel different:

• Shared identity. Everyone in a given channel accesses a single Claude identity. Anyone can see what Claude has been working on and pick up the conversation from where the last person left off.
• Persistent memory. Claude retains context about the channels work over time instead of starting cold with every mention.
• Ambient mode. Beyond responding when tagged, Claude can proactively jump into the chat on its own to keep the team updated, flag things from across the organization, and follow up on threads or tasks that were forgotten.

How it works under the hood

When you assign Claude Tag a specific task, it breaks the task into stages and works through them using whatever tools it has access to, then responds in a Slack thread with what it produced. So the unit of work is a thread, and the output lands in public view where teammates can see and continue it.

Access is gated by administrators. System admins specify which tools, information, and channels each Claude can reach. Each Claude identity stays scoped to the channels the admins define. The example Anthropic gives is pointed: a Claude set up for legal work cannot seed memories into the engineering channel.

That scoping is the mechanism that turns a single always-on AI into many purpose-built ones, each with its own walled memory.

Why this is really a fight over context

Anthropic says the result feels like working with a real colleague, one that can produce work in public view with far greater context and understanding than before. Strip away the framing and the strategic move is about capturing organizational context, institutional knowledge, and enterprise workflows.

That context is becoming the contested ground in enterprise AI, and Anthropic is not alone in chasing it. Microsoft has Graph, surfaced through Copilot and Work IQ. Snowflake and Databricks are positioning their platforms as the back end that holds tacit organizational knowledge for agents to tap. Glean is building an intelligence layer that understands company context and sits between the model and enterprise data.

The pattern is clear: whoever holds the richest, most current map of how a company actually works owns the most valuable position in the stack. An AI that quietly reads your Slack all day is one of the fastest ways to build that map.

If you deploy Claude Tag, treat channel access as a security decision, not a convenience setting. An always-on AI with read access and persistent memory inherits the sensitivity of every channel it can see. Scope each identity tightly and review those permissions before, not after, a legal or HR channel gets exposed.

What happens next over the coming 24 to 72 hours

This is a research preview, so the immediate window is about access and reaction rather than mass rollout. Expect a few things in the short term.

• Enterprise and Team admins start testing. Because the preview is limited to Claude Enterprise and Claude Team customers, the first real signal will come from how administrators configure channel scoping and which tools they grant.
• The privacy conversation sharpens. An always-on AI that learns from Slack messages and can read across channels with permission will draw scrutiny over data handling, memory retention, and who can see what Claude has learned.
• Competitors respond. With Microsoft, Snowflake, Databricks, and Glean already pushing on enterprise context, watch for positioning and feature messaging aimed at the same buyers.

What to do if you administer a Claude workspace

1. Decide which channels genuinely benefit from persistent AI memory before enabling anything.
2. Create separate, narrowly scoped Claude identities for sensitive functions like legal, HR, and finance.
3. Audit which tools and cross-channel read permissions each identity is granted.
4. Tell your team that Claude works in public view, so anything in a Claude-enabled channel is part of its working context.

Claude Tag is a small feature with a large ambition. The tagging is the surface. The persistent, organization-aware memory underneath is the actual product, and the reason this launch matters beyond Slack.

Source: TechCrunch