Anthropic’s Claude Tag Slack integration is now in research preview, positioning itself as a persistent AI teammate that lives inside your workspace, accumulates context, and acts more like a colleague than a chatbot. The pitch is familiar; the execution is a step further than what Slack’s existing AI tools have attempted.
How the Claude Tag Slack Integration Actually Works
Users already had ways to invoke Claude inside Slack: direct messages, channel @mentions for on-demand help, and Claude Code in Slack for routing coding tasks to full sessions on the web. Claude Tag replaces the existing Claude in Slack application entirely, according to TechTimes, and brings something the earlier tools lacked: continuous, shared memory across a channel.
Anyone in a given channel works with a single Claude identity. Pick up where the last person left off, see what Claude has already produced, and tag it for the next piece. Anthropic describes the aim as working with a ‘real colleague, one that can produce work in public view, with far greater context and understanding than before.’
There is also an ambient mode, which does not wait to be tagged. Claude Tag can proactively jump into conversations, flag items from elsewhere in the organisation, and follow up on threads that have gone quiet. Whether teams will want an AI interjecting unsolicited is a reasonable question; the feature can presumably be configured by administrators.
Speaking of administrators: setup involves pairing Claude Tag with a Slack workspace, granting it access to specific tools and data sources, setting spending limits, and defining which channels it can operate in, VentureBeat reports. A Claude instance scoped to legal cannot bleed memories into the engineering channel. That scoping matters: enterprise buyers have been burned by AI tools that treat the entire company’s data as a single undifferentiated pool.
Under the hood, Claude Tag holds its own service accounts in each connected system, posting in Slack as the Claude app, opening pull requests as the Claude GitHub App, and querying connected data warehouses under an administrator-provisioned service account.
The Multiplayer Argument, Backed by Internal Numbers
Cat Wu, Anthropic’s head of product for Claude Code and Cowork, told Fortune that the distinction from Anthropic’s previous tools is structural: ‘Claude Code, CoWork, and chat are very single player, whereas Claude Tag is built to be interactive and multiplayer.’
Wu also offered a number that Anthropic will inevitably repeat in every enterprise sales conversation: within Anthropic itself, Claude Tag is already approving and incorporating 65% of the code changes the product team submits. Take that figure with the usual caveat that self-reported internal metrics are not exactly independent audits, but it is a concrete data point in a category full of vague productivity claims.
Claude Tag launches for Claude Enterprise and Claude Team customers in research preview. Engadget reports that Anthropic expects to expand availability beyond the research preview stage, though no timeline has been specified.
Slack Is the Start, Not the Finish
Anthropic is eyeing Microsoft Teams, email, and project management tools as future destinations for Claude Tag, so that teams can tag @Claude across the platforms where they already work, according to VentureBeat. That ambition puts it squarely in competition with Microsoft’s own Copilot, which surfaces organisational context through Microsoft Graph, as well as Glean’s intelligence layer that sits between the model and enterprise data.
The Salesforce angle is worth watching. As part of an expanded partnership, Anthropic and Salesforce announced that Salesforce is deploying Claude Code across its global engineering organisation, Anthropic is deepening its use of Slack, and Claude is becoming a preferred model for Salesforce’s Agentforce platform in regulated industries including financial services, healthcare, cybersecurity, and life sciences. A tighter Slack relationship is now commercially baked in.
The Claude Tag Slack integration is, at its core, a bet that persistent organisational context is the feature enterprise customers will actually pay for, not just raw model capability. The 65% internal code-approval figure is doing a lot of work in that pitch. Whether external teams, with messier data and less controlled environments, see comparable results will be the real test when Claude Tag moves out of research preview.
