Anthropic Launches Claude Tag in Slack Research Preview as an Always-On AI Teammate
Anthropic launches Claude Tag in Slack research preview, an always-on AI teammate for Enterprise and Team customers that preserves channel memory and more.
Anthropic has begun a research preview of Claude Tag, a persistent “always-on” Claude that lives inside Slack and functions as an AI teammate for collaborative channels. Claude Tag is designed to maintain ongoing context and shared memory across defined channels, enabling team members to tag @Claude for insights, task assignments, and follow-ups. The feature will be available to Slack users on Claude Enterprise and Claude Team plans during the research preview phase.
Product rollout and availability
Anthropic said Claude Tag will enter a research preview initially for Slack users on corporate Claude plans, giving enterprises early access to the persistent assistant. The preview limits participation to customers on Claude Enterprise and Claude Team tiers and runs as an integration inside Slack rather than a separate application. Anthropic frames the release as an evolution of its existing Slack integrations and a step toward deeper, continuous assistant behaviors within workspaces. The company will gather feedback during the preview to refine memory, permissions, and workflow behavior.
How Claude Tag preserves channel memory
Claude Tag is built to maintain context from channel activity over time, so the assistant “learns” about ongoing projects as conversations evolve. That persistent context allows Claude Tag to pick up where a colleague left off and present what it has already been tracking in public threads. Administrators can scope where Claude Tag stores and retrieves context so that memories remain confined to the channels they designate. This scoped memory model aims to keep project knowledge accessible within teams while preventing cross-team contamination.
Task orchestration and ambient assistance
When assigned work, Claude Tag will decompose tasks into sequential stages and execute them using the tools it has permission to access, reporting progress back in Slack threads. The assistant can post deliverables and status updates directly into conversations so teammates can review output without leaving Slack. Claude Tag also runs an ambient mode that proactively interjects to flag stale threads, follow up on unfinished tasks, or surface relevant organization-wide information. That mix of reactive and proactive behavior is intended to make the assistant feel like a publicly visible colleague rather than a private, ephemeral helper.
Permissions, admin controls and data boundaries
System administrators will set explicit permissions for each Claude identity, determining which channels, tools and data sources it may read or write to. Anthropic emphasizes that every Claude identity is scoped to the channels administrators define, preventing a Claude configured for legal work from leaking memories into engineering discussions. Organizations can also control whether Claude Tag may pull facts from other channels or systems, enabling a balance between helpful cross-referencing and strict data governance. These controls are central to enterprise deployment scenarios where confidentiality and regulatory compliance are priorities.
Integration with existing Claude features and developer workflows
Claude Tag represents a next step beyond existing Slack functionality that lets users DM or mention @Claude for on-demand help. Earlier integrations already route coding requests into focused web sessions and return updates to Slack threads, and Claude Tag builds on that by adding continuity and shared identities. The assistant is intended to coordinate multi-stage tasks across available tools, integrating with coding, document and data workflows where organizations permit access. For teams that already use Claude for code and content, the Tag model aims to reduce repetitive context-setting and handoffs.
Enterprise context and market positioning
Anthropic positions Claude Tag as a solution for capturing tacit organizational knowledge that agents need to operate effectively in enterprise settings. Competitors and partners are also focusing on context and knowledge layers: Microsoft offers Graph-driven features through Copilot, while vendors such as Snowflake, Databricks and Glean promote platforms that surface company data and context for agents to use. Anthropic’s approach centers on a shared Claude identity inside collaboration tools, making the assistant’s activity visible to team members while relying on administrative controls for safety.
Research preview expectations and next steps
During the research preview, Anthropic will monitor how Claude Tag handles ongoing context, permission scopes and ambient interventions and will tune behaviors based on customer feedback. Enterprises participating in the preview can experiment with channel-scoped memories, task orchestration and cross-channel data access under administrator supervision. The research phase will inform any broader rollouts, additional integrations and policy controls that enterprises expect from assistant deployments. Observers say the preview will be a key test of whether always-on assistants can deliver value without creating governance or privacy issues.
Anthropic’s Claude Tag trial in Slack reflects a growing industry focus on persistent context and collaborative agents, aiming to reduce repetitive context-setting and make AI contributions visible within team workflows while leaving control of data and access in administrators’ hands.
