LangSmith Fleet: Enterprise Agent Management

LangChain has launched LangSmith Fleet, an evolution of its Agent Builder that transforms agent development from a single-person tool into an enterprise-wide platform. The rebranding reflects a shift in focus: while Agent Builder emphasized creation, Fleet emphasizes governance, sharing, and operational oversight as organizations scale from a handful of agents to dozens deployed across teams.

Core Features

Permissions and Sharing: Fleet introduces granular access controls across three permission levels—"Can clone" (customize for personal use), "Can run" (execute without modification), and "Can edit" (full configuration access). Teams can assign these permissions to individual users or entire workspaces, enabling secure agent distribution while preventing unauthorized modifications.

Agent Identity Models: The platform now supports two authentication patterns. "Claws" agents use fixed credentials regardless of who invokes them, ideal for team tools like shared Slack bots. "Assistant" agents authenticate on behalf of the invoking user via OAuth, maintaining individual permissions and audit trails—useful for personal-access workflows like document retrieval.

Agent Inbox: A centralized dashboard for human-in-the-loop oversight, allowing managers or admins to review, approve, or reject agent actions across the entire organization without context switching.

Slack Bot Integration: Each agent can now have its own Slack bot handle (@vendor-intake, @weekly-sales-numbers), enabling direct team interaction while maintaining separate credentials and permissions.

Enterprise Observability

Fleet builds on LangSmith's existing tracing infrastructure, providing a complete audit trail: which agent acted, on whose behalf, with what credentials, and every decision made. This addresses critical compliance and debugging needs for enterprise deployments.

Next Steps

LangChain indicates rapid iteration ahead, with additional features for sharing, identity, and autonomous work coming in the coming weeks. Organizations managing multiple AI agents should evaluate whether Fleet's permission model and observability align with their governance requirements.