Anthropic introduces the agent identity access model for Claude Tag, its multiplayer AI product for team workspaces. Instead of acting on behalf of individual users, Claude operates under its own service accounts provisioned by admins. Permissions are scoped at the workspace and channel level rather than per-user, solving the problem of whose credentials apply when multiple team members interact with the same agent. Admins control repository access, API connectors, skills, and standing instructions per channel. Private channels get distinct identities; public channels share a workspace-level identity. Every action is logged under Claude's own service accounts. Future plans include just-in-time credential grants and identity-aware overlays for complex clearance structures.
