Managing AI Agents at Scale

LangChain has launched LangSmith Fleet, a major update to their LangSmith Agent Builder that transforms individual agent creation into enterprise-grade fleet management. The platform enables teams to create agents from natural language prompts, then share, control, and monitor them across the organization with granular permissions and full observability.

Key Features

Flexible Sharing and Permissions: Agents can be shared with individual users or entire workspaces with three permission levels—clone (customize), run (execute only), or edit (full control). These permissions can be layered across different user groups and adjusted at any time.

Agent Identity and Credentials: Two authentication models provide flexibility:

• "Claws": Fixed credentials that work for any user running the agent, ideal for shared team resources like a Linear Slack bot
• "Assistants": User-based authentication where each person authenticates with their own credentials, maintaining individual access controls

Slack Bot Integration: Each agent can now have its own Slack bot identity, allowing teams to @mention specific agents (@vendor-intake, @weekly-sales-numbers) instead of routing everything through a single bot.

Agent Inbox and Human-in-the-Loop: A centralized inbox provides oversight across all running agents. Users can review, approve, or reject agent actions from a single dashboard. For "Claws," edit-access users see all activity; for "Assistants," each user's actions remain private.

Enterprise Observability

LangSmith Fleet builds on LangSmith's native tracing capabilities, capturing every tool call, decision, and output in a structured, searchable audit trail. Combined with agent identity and permissions, this creates a complete audit record showing which agent acted, on whose behalf, with what credentials, and what happened at each step.

What Developers Need to Know

This release is particularly valuable for organizations building multi-agent workflows. Teams can now securely publish agents for broader use, enforce compliance through audit trails, and manage agent authentication across multiple tools and users. The platform transitions from individual engineer-driven agent creation to knowledge workers building and sharing agents across the organization.