Meet LangSmith Fleet: A Non-Technical Agent Builder That Lets Anyone Build, Use, and Manage AI Agents

Autonomous AI agents are already being built and deployed by AI and tech companies; however, building an AI agent isn't an easy task, even the most experienced developers need multiple frameworks to build a reliable agent. So, what if your entire team of technical and non-technical members has the ability to build, use, and manage an agent fleet with the security your organisation requires? LangChain has launched LangSmith Fleet, which allows your team to do exactly that.

From Agent Builder to Fleet: What Changed and Why

LangSmith Fleet, formerly LangSmith Agent Builder, is an enterprise workspace for creating, using, and managing a fleet of agents. Each agent has its own memory, access to a collection of tools and skills, and can be exposed through the communication channels teams use every day.

LangChain launched Agent Builder in October to allow knowledge workers to create their own agents using natural language.

LangChain said that they have seen teams consistently start with one or two agents for simple tasks like research or status checks, then expand to running more tasks across more agents. As that scale grew, questions around ownership, authentication, and auditability became the hard part. Fleet is LangChain's structured answer to that.

Tiered Permissions That Mirror Real Team Structures

One of Fleet's most practical additions is a sharing and permissions model, configurable per agent with two dimensions, which are who gets access and what they can do. Three permission levels govern this:

• Can clone: Copy the agent into a personal, customizable version.
• Can run: Use the agent without modifying its configuration.
• Can edit: Full access to change instructions, tools, and settings.

These can be layered as needed, for example, giving a core team edit access while sharing run-only with the broader workspace, and permissions can be changed or revoked at any time.

Agent Identity: Claws vs. Assistants

Perhaps the most technically significant feature is Fleet's dual-identity credential model for controlling how agents authenticate with external tools:

• Claws: Claws use a fixed set of credentials regardless of who runs them, so users don't need to log in to each tool. This is useful for something like a Linear Slack bot, where the entire team can search for and create issues using the same credentials.
• Assistants: Assistants act on behalf of the user who invokes them. Each user authenticates with their own account for each connected tool via OAuth. This makes sense for something like a team knowledge base in Notion, where each user has different access to documents.

Custom Slack Bots Per Agent

Fleet agents can already respond to messages in Slack. With agent identity, each agent can now have its own Slack bot triggered by its defined name. Teams can create a bot for each job and assign it a Slack handle: @vendor-intake, @weekly-sales-numbers, @onboarding-agent, allowing team members to @mention an agent in a channel or DM it directly without switching context.

Agent Inbox and Observability

The Agent Inbox provides a single place to review, approve, or reject actions across all running agents, without switching across tabs, supporting human-in-the-loop oversight for both Claws and Assistants based on specific permissions.

On the audit side, LangSmith already provides native tracing for every agent action in Fleet. Every tool call, every decision, and every output is captured in a structured trace that can be inspected, searched, and exported, giving enterprises a complete picture of which agent acted, on whose behalf, with what credentials, and what it did at each step.

In Conclusion:

LangSmith Fleet shows a broader maturation happening across the AI agent ecosystem. The tooling conversation has changed from "how do we build agents?" to "how do we run them responsibly at scale?" Now, anyone on your team can build a custom AI agent while keeping the human in the loop and maintaining full transparency.

LangChain says it is actively expanding Fleet, with more coming soon for agent sharing, identity, and safe autonomous work in the weeks ahead.