Anthropic Claude Tag Launches For Slack To Enable Autonomous Project Planning And Complex Team Workflows
Anthropic has rolled out Claude Tag which involves its assistant being plugged directly into Slack channels as a working participant in the team. This feature enables teams to assign complex work to the model by mentioning it directly in shared workspaces. As summarized by the official Anthropic product documentation, it is now in a beta release for Enterprise and Team tier subscribers taking over the venerable Slack app.
This system is so useful that internally at Anthropic, the product group uses it for daily work, according to internal metrics. Company data shows that approximately 65% of code written by their internal product group is also written by Claude Tag. Aside from normal software engineering, it is also being used at teams to track product performance metrics, handle customer support tickets, and troubleshoot hard to identify software bugs.
No typical single user ChatLike, Claude Tag works on a multiplayer channel. One instance of the assistant converses with the entire workspace, and anyone can see the thread what questions have been asked and what progress has been made; any teammate can join the session.
The system is meant to build context knowledge over time along with the channel conversations, so the customer wouldn't need to keep messaging details of the projects over and over again. If given an admin permission, the assistant will be able to also collect information from other slack channels as well as from connected databases, while failing to access to private channels. This knowledge will be used by the model to provide more accurate and relevant results based on the workspaces specific history.
When ambient mode is enabled the assistant can also be more proactive it can actively book marks critical announcements across connected channels, track active projects, surface essential information and followup on inactive, silent threads. Its asynchronous architecture also allows teams to hands off multi stage assignments of greater complexity which the assistant can independently plan, schedule and implement over hours or days while other team members are working on their tasks.
Admin has tight control over data privacy and Tool access. Admin can instantiate multiple AIs for different Use Channels for example gives sandboxed environment for sensitive data. An assistant built for sales workflows cannot share its memory or toolset to a channel for engineers. Admin also has the capability to impose token spend limits at Organization and Use Channel levels and has a detailed activity log of every task and agent.
Software is built on the Opus 4.8 engine. Administrators can select to migrate from the legacy app using their console within 30 days. Anthropic is also providing discounted launch credits for eligible Enterprise and Team organizations to enable testing in private channels ahead of launch.
