Schedules for Managed Deep Agents: Cron jobs, prompts, and Slack delivery

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Schedules for Managed Deep Agents: Cron jobs, prompts, and Slack delivery · Channel: LangChain · 03:17 · Watch on YouTube

Topics: AI agents, Task and workflow automation

Summary

The video explains how to configure recurring schedules for managed agents using cron files, prompts, time zones, and optional input context. It demonstrates setting up a weekly Salesforce pipeline summary and delivering it to a configured Slack conversation. After deployment to LangSmith, the schedule and prompt are visible in the deployment UI, and each run creates a thread and posts its result to Slack. Scheduled agents can also be used as standard agents, and their tool use can be inspected in LangSmith traces.

Key points

Schedules let managed agents send recurring messages that use their tools and available contacts.
A schedule is defined in a file in the schedules directory with a cron, time zone, and prompt.
Optional schedule input can provide context beyond the basic prompt.
The example configures a weekly Monday pipeline summary using Salesforce data.
A configured Slack channel directs the generated summary to a specific conversation.
Deploying the agent to LangSmith publishes it and configures its cron schedule.
The LangSmith deployment UI shows the agent’s schedules and their prompts.
Each scheduled run creates a thread and can publish its message to Slack.
An agent configured with a schedule can still be used as a standard agent.

Chapters

Start	Chapter	Summary
00:00	Introducing managed agent schedules	The presenter introduces recurring messages and describes using a schedule to send a weekly sales pipeline summary.
00:32	Configuring a schedule and prompt	The example agent has Salesforce tools, and the schedule is defined with a cron, time zone, prompt, and optional context.
01:41	Deploying the cron to LangSmith	The agent is deployed, and the LangSmith UI shows the configured cron and prompt.
02:15	Viewing the Slack result	The scheduled run creates a thread, uses the agent’s tools to access Salesforce data, and publishes the summary to Slack.
02:46	Using the agent beyond its schedule	The presenter shows that a scheduled agent can also answer questions as a standard agent and concludes the introduction.

Quotes

Schedules can be configured and added to your agent — 00:00

You’re going to create a new file in the schedules directory — 00:32

this will publish the agent to LangSmith, and it’ll configure the cron. — 01:41

doesn’t mean that’s the only thing that it can do. — 02:46

Concepts and entities

Managed Deep Agents, schedules, cron, MCP, LangChain, LangSmith, Salesforce, Slack, Nathan

Full transcript (original language)

00:00 Hey, I’m Nathan. I’m a product here at LangChain, and today we’ll be talking about schedules for managed agents. Schedules can be configured and added to your agent to allow it to send recurring or regular messages that invoke its tools or leverage contacts that it has available. You can use a schedule to, for example, send a weekly summary of GitHub PRs or send a summary of tickets that have been opened inside of your ticket triage system. In this example, we’re going to use schedules to send a pipeline summary. So, every week the sales team can see updates to the pipeline and deals that may be at risk.

00:32 To start, we have our agent that’s configured, and this agent is set up with an instructions file, which you can see here. This agent has some tools that represent access to Salesforce. Now, these are a mock connection, but in real life you would connect to Salesforce using an MCP server, for example. If we take a look at the schedule system, the syntax is pretty straightforward. You’re going to create a new file in the schedules directory, and you can define the schedule using a cron at specific time zone and with a specific prompt. You can also configure input to the schedule to provide context for it

01:07 beyond a basic prompt, if that’s something that you need. For example, a user message. So, the first thing we’re going to do is set up our base cron, where we send a pipeline summary every Monday based on data that the agent has access to from Salesforce. So, we have our prompt, right? The Monday pipeline for the sales team and the specific structure that we want this written in. We also want to make sure that this is delivered to Slack. We’ve configured a Slack channel for this agent using the channel system, which you can take a look at another video to learn more about that. In this situation, this will ensure that every time this summary is generated, it’s actually delivered to a

01:41 specific conversation. Now that we have our cron set up in our managed deep agent, we’ll want to deploy this to LangSmith. So, I can do so with a deploy command, and when I run this, this will publish the agent to LangSmith, and it’ll configure the cron. So, let’s let this So, finish deploying with the new cron, and if we take a look in LangSmith, we can actually see the setup here in the UI. So, under our deployment for our managed deep agent, we can see in the crons tab, we have the set of crons that are associated with this agent, one per file. Here’s the schedule, and here’s the prompt that we

02:15 set in our agent source. When this agent runs, it’ll create a new thread associated with the agent server, and it’ll also publish the message to Slack. Well, here it is in the channel that we’ve configured. So, based on this prompt that we can see here, write the Monday pipeline summary, leak with risk and out volume, um and a specific structured output, it will do that using the set of tools and metadata that it has available. So, in this case, it’s able to interact with Salesforce, so it can pull real data out of our Salesforce database. So, bear in mind, of course, any agent that you have

02:46 that’s set up with a schedule can still be used as a standard agent. So, interact with this as normal by tagging it directly, and ask it a question that it has data about, it can answer that question as well. So, just because an agent is configured on a schedule, doesn’t mean that’s the only thing that it can do. So, I look in this thread, it’s able to use its tool calling and actually fetch this data. And as always, I can check out the details of this trace in LangSmith. So, that’s a quick introduction to schedules for managed deep agents, which you can use to configure automated

03:17 recurring messages from your agent into Slack or other channels. Take it for a spin, start building out some automated systems that are powered by your agent.