Scheduled Agents

Claude on Autopilot

Up to now, everything in this course has been interactive. You type a prompt, Claude does work, you review the result. That is powerful, but it still requires you to be there, driving.

Scheduled agents change the equation. They run Claude Code automatically, on a timer, in the background, without you touching a keyboard. This is the shift from "I use Claude when I have a question" to "Claude handles this for me every morning."

What a Scheduled Agent Is

Think of it like a cron job (Lesson 5), but instead of running a simple script, it runs Claude Code with a specific prompt. Every morning at 8 AM. Every Monday afternoon. Every night at midnight.

The scheduled agent wakes up, connects to everything through the harness (Lesson 9), executes a skill or a prompt, and goes back to sleep. You do not need to be at your computer. You do not even need your computer to be on if you run it remotely.

A scheduled agent might:

• Pull yesterday's data from three different APIs, generate a summary briefing, and send it to your inbox
• Scan your project for security vulnerabilities every night and open a ticket if it finds something
• Check your project board, update overdue items, and send you a status digest
• Generate invoices from tracked hours and email them to clients

Agents vs. Skills

A skill is a recipe. A scheduled agent is the chef who follows the recipe on a schedule.

You can run a skill manually any time by typing its slash command. A scheduled agent just automates that invocation so it happens at a specific time without you. The agent uses the same harness, the same MCPs, the same CLAUDE.md. The only difference is who presses the button: you or a timer.

You can also run agents that are not tied to a skill. A scheduled agent can receive any prompt, not just a skill name. "Check if anything urgent came in overnight and text me if so" is a valid agent prompt. But in practice, the most reliable agents run well-tested skills, because the workflow is predictable and repeatable.

The Full Chain

Everything in this course connects. Here is the complete workflow:

1. Create a project. Set up a folder, open it in your editor.

2. Run /init. Give Claude persistent memory about the project (Lesson 4).

3. Install MCP servers. Connect Claude to your external tools: project management, email, databases, whatever you need (Lesson 8).

4. Pull real data. Ask Claude to query your actual systems. "Show me last week's open tickets." "Pull all contacts who signed up this month." Real data, live systems, instant results.

5. Build something. "Create a dashboard that shows this data." "Write a service that generates a weekly summary email." Claude writes the code, using real data as reference.

6. Create a skill. Encode the workflow so you can repeat it with a single command (Lesson 10).

7. Schedule it. Set up a scheduled agent so it runs automatically. Every Monday morning, the summary generates and sends itself. You built it once; it runs forever.

That chain (create, init, connect, query, build, skill, schedule) is the full speed stack. Each step uses concepts from a specific lesson. Together, they produce a working, automated system.

What You Can Do Now

You can now do four things that most people cannot:

1. Read architecture. When someone explains a system, you understand the pieces. APIs, databases, triggers, containers, MCP. You know what they do.

2. Describe systems. You can take a business need ("send the team a daily briefing") and break it into components: a cron trigger, MCP-connected APIs, a database query, an AI call, an email send.

3. Direct AI to build. That description is exactly what you give to Claude Code. With a CLAUDE.md providing context and MCPs connecting to your tools, Claude turns plain English into working software.

4. Automate with agents. Once it is built, you schedule it to run itself. A daily briefing, a weekly report, a nightly cleanup. Claude does not just build it, it keeps running it.

You do not need to become an engineer. You have learned something arguably more valuable for your role: the ability to understand, describe, direct, and automate the systems that run your work.

Further Reading