Anthropic moves AI agents from private chats into shared Slack channels

The integration, currently in beta for Enterprise and Team tiers, marks a departure from traditional browser-based AI workflows. By embedding the Opus 4.8 engine into Slack, Anthropic enables multiple employees to monitor the model’s progress in real-time. This shared visibility reduces the need for constant context-switching, as the AI maintains an automated history of project scopes and foundational company data.

Beyond basic interaction, the system features an "ambient" configuration that allows agents to function asynchronously. Once activated, these agents can monitor threads, track unresolved assignments, and process incoming emails without requiring continuous human prompts. According to internal metrics, this approach has gained traction, with Anthropic’s product group reporting that 65% of its code is now generated through these internal Claude agents.

However, the transition to persistent, channel-based automation introduces new security hurdles. Granting AI agents access to chat histories and code repositories necessitates granular control, prompting Anthropic to introduce administrative portals for managing data access and monitoring token costs. While the shift promises to lower task friction and streamline routine knowledge work, it leaves IT departments responsible for rigorous auditing to prevent sensitive information from crossing into unauthorized channels.