24 lessons, 0 hours of video: LangChain's new course covers how agents go live

LangChain has launched a new course on Academy called Building and Deploying Managed Agents. It is free, and the instructor is Ishaan Lal.

The reason given on the course page is very specific: getting an agent from working to deployable can take several weeks, and the time is not spent on the model but on the surrounding infrastructure that has to keep up. This course aims to close that runtime gap.

The course has 24 lessons divided into four modules.

Module one covers how to build an agent: an overview of Managed Agents, the first agent, Context, Tools and Middleware, Subagents, and finally a lab.

Module two moves to the runtime side: Agent Server, Sandboxes, Connections, and Schedules, all things you only run into once an agent is running.

Module three addresses the question of who invokes it: Identity, the concept of Channels, Slack Channels, HTTP Channels, and building your own frontend, again with a lab.

Module four has only one lesson: Evals. According to the course page, these evals are meant to catch problems early rather than waiting for users to report them.

One detail is worth noting. The course page lists 24 lessons and 0 hours of video content. Twenty-four lessons, zero video hours. Either the videos are still being uploaded, or the course is simply not centered on video. Anyone expecting a recorded lecture course should set expectations accordingly.

The course revolves around LangSmith Managed Agents, offering a managed runtime, connections, user authentication, and evals. You can register at academy.langchain.com/courses/managed-agents. It is free.

For developers who are already writing agents and are stuck at the deployment stage, the lab portions of the course may be more useful as a reference than the syllabus.