Sixty-five percent of Anthropic’s own product team’s code now gets generated through an internal agentic Slack workflow. That figure, disclosed by Anthropic in its announcement of Claude Tag, is the clearest signal yet that the company is betting on ambient, team-wide AI participation rather than individual chat sessions as the default mode of work.
Claude Tag is a Slack integration that lets any member of a channel assign tasks to Claude by mentioning @Claude directly. Unlike single-session tools where each conversation starts cold, Claude Tag maintains memory scoped to its assigned channels. It builds context as conversations happen around it, retaining what it needs to perform follow-up tasks without requiring users to re-brief it each time.
The architecture has three features that separate it from prior Slack bots and even from Claude’s own existing integrations. First, a single Claude instance per channel means every team member interacts with the same persistent agent, not isolated per-user sessions. Anyone can pick up a task thread where someone else left off. Second, ambient mode lets Claude proactively surface updates and flag stalled threads without waiting to be tagged. Third, Claude can schedule work autonomously across hours or days, running tasks in parallel for multiple requesters.
Anthropic positions Claude Tag as an evolution of Claude Code, its existing autonomous coding product, made collaborative. Where Claude Code handles individual developer tasks, Claude Tag is designed for organizational workflow: the announcement describes use cases including product metrics analysis, support ticket processing, and root-cause investigation, not just code generation.
The access model matters here. Claude Tag replaces Anthropic’s existing Claude in Slack app, and Anthropic is offering a launch credit to eligible Enterprise and Team organizations to lower the cost of initial adoption. Administrators control access channel-by-channel, with separate memory scopes per channel so a Claude set up for sales cannot share memories or data access with one configured for engineering. Token spend limits are configurable at the organization and channel level, and a full activity log is available. This is an access model designed for legal and compliance teams to approve, not just engineers.
The critical question for builders evaluating this is whether persistent channel memory holds up over weeks of high-volume usage. Anthropic has not disclosed how memory is retained, compressed, or bounded. Products that rely on Claude Tag for workflows that span months will need to test degradation behavior before treating it as a production dependency.
The competitive framing is worth noting. Microsoft 365 Copilot has offered persistent Slack and Teams integrations for some time, and OpenAI’s GPT integrations have appeared inside Slack through various third-party channels. Claude Tag’s differentiation is the explicit multi-user collaboration layer and the asynchronous task scheduling, which goes further than most existing Slack AI integrations in treating the channel itself as the unit of work rather than the individual conversation.
Claude Tag launches in beta for Claude Enterprise and Team customers today, running on Opus 4.8. Administrators have a 30-day window to migrate from the legacy Slack app.
Teams building internal tooling on top of Claude should evaluate whether Claude Tag’s channel-scoped memory model fits their access control requirements before committing to the migration, particularly in organizations where data separation between departments is a compliance requirement, not a preference.
Announced by Anthropic on June 24, 2026, at anthropic.com/news/introducing-claude-tag.
