LangChain Launches LangSmith Fleet for Enterprise AI Agent Management
Timothy Morano Mar 19, 2026 17:30
LangChain rebrands Agent Builder as Fleet, adding enterprise controls for permissions, credentials, and multi-agent oversight across organizations.
LangChain has rebranded its Agent Builder product as LangSmith Fleet, expanding the no-code agent platform with enterprise-grade permissions, credential management, and centralized oversight tools. The update addresses a growing pain point: organizations that started with one or two AI agents now find themselves managing dozens without proper governance.
The rebrand reflects a shift in how companies are deploying AI agents. Six months ago, building an agent required engineering resources. Now anyone can describe a task in natural language and spin up an agent to handle it. That democratization created a new problem—agent sprawl.
What Fleet Actually Does
Fleet introduces tiered permissions controlling who can edit, run, or clone each agent. Workspace admins can publish agents company-wide while restricting modification rights to core teams. The system supports two distinct authentication models:
Claws operate with fixed service credentials regardless of who triggers them. Useful for shared resources like a Linear integration where the whole team creates and searches issues through one account.
Assistants authenticate as the individual user invoking them. Each person logs into connected tools with their own credentials, and the agent inherits their specific permissions. Makes sense for Notion knowledge bases where document access varies by role.
Slack Gets Native Agent Handles
Fleet agents can now have dedicated Slack identities instead of routing everything through a single bot. Teams can create handles like @vendor-intake or @weekly-sales-numbers, letting employees hand off tasks without leaving their conversation flow.
The Inbox feature provides human-in-the-loop oversight across all running agents from one dashboard. For Claws, only users with edit access see the action queue. For Assistants, each user's activity stays private to their own Inbox—important when agents handle sensitive documents.
The Audit Trail Angle
Every tool call, decision, and output gets captured in structured traces that can be searched and exported. Combined with the identity and permissions layers, enterprises get a complete record: which agent acted, on whose behalf, with what credentials, and what happened at each step.
LangChain says additional agent identity features and expanded sharing controls are coming in the weeks ahead. The platform is available now at smith.langchain.com.
Image source: Shutterstock.jpg)