Module 7 of the Hermes Agent Masterclass explains how to use Cron for scheduling autonomous, repeatable tasks in isolated sessions, with flexible creation methods, customizable output delivery, and cost-saving strategies like running scripts without language models. It also covers chaining Cron jobs into multi-stage pipelines for efficient workflows and previews the upcoming topic on sub-agents for parallel automation.
In Module 7 of the Hermes Agent Masterclass, the focus is on Cron and automation, highlighting how Hermes Agent enables scheduled, autonomous task execution without user intervention. Unlike previous modules where user input was necessary to trigger agent actions, Cron allows Hermes to run tasks on a schedule in fresh, isolated sessions, delivering results directly to the user’s preferred platform, such as Telegram. The module covers how Cron operates within the Hermes gateway, which ticks every minute to check and execute due jobs, emphasizing the importance of keeping the gateway running for Cron jobs to function reliably.
There are four main ways to create Cron jobs in Hermes: via the command-line interface (CLI), through natural language commands in chat (e.g., Telegram), using the text user interface (TUI), and through the Hermes web dashboard. The dashboard also offers useful blueprints for common tasks like morning briefings and reminders, making setup straightforward. Hermes includes safety features to prevent runaway job creation loops by disabling scheduling tools during Cron executions. Scheduling supports natural language intervals, fixed durations, and traditional cron expressions, providing flexibility for various automation needs.
Delivery of Cron job outputs is highly customizable, allowing messages to be sent back to the origin platform, saved locally, or delivered to multiple messaging channels simultaneously. Hermes supports a “silent” mode to suppress output when there is nothing noteworthy, which is particularly useful for monitoring tasks. A key consideration is that each Cron job runs in a fresh session with no memory of previous runs, so prompts must be fully self-contained. To provide context or state across runs, Hermes offers tools like pre-run scripts, work directories, and the “context from” feature, which pipes output from one job into another.
Cost-saving strategies are a major highlight, with Hermes enabling zero-cost Cron jobs by running simple scripts without invoking language models (“no agent” mode) and using a “wake agent” gate that only activates the model when necessary based on script output. This approach allows frequent polling or monitoring without incurring model usage costs unless there is a meaningful event to process. Additional optimizations include pushing data fetching to free scripts and limiting enabled toolsets per job to reduce token usage. Hermes also ensures resilience by inheriting fallback providers and credential rotation from user settings, maintaining reliability under rate limits or errors.
Finally, the module demonstrates how to chain Cron jobs into pipelines using the “context from” feature, enabling multi-stage workflows without complex orchestration frameworks. For example, one job can collect data hourly with no agent, and a second job can analyze and summarize that data daily, delivering a concise report to Telegram. This setup reduces errors by handling data handoff outside the language model and supports fan-in from multiple upstream jobs. The module concludes by previewing the next topic on sub-agents, which expands automation by running multiple agents in parallel, complementing Cron’s time-based scheduling for more powerful autonomous workflows.