Anthropic Launches Claude Tag for Slack, Claims 65% of Internal Code Created With It
Claude Tag introduces a new access model where the AI operates as a persistent team member with its own permissions, moving beyond individual chat assistants toward autonomous, team-wide agents.
Reporting from 2 sources: GameBusiness.jp, GIGAZINE.
Anthropic has announced Claude Tag, a new feature that integrates the Claude AI assistant into Slack as a team member. Users can delegate tasks by tagging @Claude in channels, where the AI reads permitted thread conversations and responds with context. Claude Tag is available in beta for Claude Enterprise and Team plan subscribers. The feature introduces four capabilities distinct from individual AI use: multiplayer support allowing anyone in a channel to interact with the same Claude instance; continuous learning from channel conversation history; proactive behavior where Claude actively notifies relevant information and follows up on unresolved threads; and asynchronous task processing that lets Claude work on projects over hours or days while users focus elsewhere. Claude Tag adopts a new access management model called Agent ID, giving the AI its own account per channel with permissions set by administrators, preventing cross-channel information leaks. Anthropic has been running an internal version and reports that 65% of code created by its product teams is generated by Claude Tag. The tool runs on the Opus 4.8 model. Migration from the existing Claude in Slack app is available via opt-in within 30 days, with onboarding credits for eligible customers.
Anthropic's internal use of Claude Tag extends beyond coding to product metrics tracking, support ticket handling, and bug root cause analysis, according to the company. The feature, announced on June 24, 2026, replaces the previous Claude in Slack app and is designed for collaborative team environments rather than individual use. Slack was chosen as the first supported application because Anthropic uses it internally, with plans to expand to other apps.
Claude Tag's Agent ID permission model gives the AI its own account per channel, separate from any user's credentials. Administrators set tool access per channel-for example, Claude in an engineering channel can reach GitHub and data warehouses, while Claude in a sales channel is limited to CRM tools. No information leaks across channels, and private channel data stays contained. In direct messages, Claude retains the same permissions as the traditional one-on-one version, with conversation content kept private per user. Administrators can set token consumption limits and view audit logs of all operations.
Eligible customers can start by linking their Slack workspace and granting tool access. The migration from the existing Claude in Slack app must be completed within 30 days via opt-in, and onboarding credits are provided. Claude Tag runs on the Opus 4.8 model.
Synthesized by Yomimono from the 2 cited sources below, including Japanese-language reporting where cited, then editorially reviewed before publishing.