CF

Hands-On Loop Engineering with Claude Course

2h 12m 35s
English
Paid
Added to CourseFlix

Hands-On Loop Engineering with Claude Course is a 9-lesson 2 hours 12 minutes self-paced course by Youssef Hosni. The Practical Course on Loop Engineering with Claude helps transition from one-time prompts to repeatable work cycles : processes that run on a schedule or event, read the necessary context, check results, save state between runs, and…

Course facts

Lessons
9
Duration
2 hours 12 minutes
Language
English
Added to CourseFlix
October 2026
Instructor
Youssef Hosni
Price
Premium

The Practical Course on Loop Engineering with Claude helps transition from one-time prompts to repeatable work cycles: processes that run on a schedule or event, read the necessary context, check results, save state between runs, and timely hand over tasks to humans.

If you write the same prompt for Claude every morning, manually check the answers, and explain the context again, the problem is not with the quality of the prompt. The issue is that the unit of work remains an individual session. For recurring engineering tasks, it's more efficient to build not an "ideal prompt," but a managed loop — a cycle around Claude Code with memory, triggers, checks, and constraints.

In the course, you'll learn from scratch how to design such cycles: what triggers the next iteration, what data Claude should read before starting work, what he's allowed to change, where the state is stored, how the result is verified, and under what conditions the loop should stop, repeat, or request human involvement.

This is a practical course lasting 2 hours and 12 minutes. The package also includes a 117-page book, 247 slides of presentation, all the prompts from the workshop, and three ready-to-use projects with working loops that you can study, adapt, and launch independently.

What is Loop Engineering in Claude

Loop Engineering is an approach to creating repeatable AI processes where Claude performs not a single instruction but part of a managed cycle. Such a cycle can analyze a project daily, respond to changes in GitHub, prepare reports, conduct reviews, update the state, and verify its own results through a separate validation stage.

Unlike simple prompting, a loop describes not only the request to the model but also the entire working system around it:

  • trigger: schedule, command, or event in the repository;
  • instruction files and context Claude needs to read;
  • state file saving memory between sessions;
  • directory or format for the work results;
  • check stage separated from the execution stage;
  • conditions for stopping, relaunching, and escalating to a human;
  • permission boundaries to ensure AI doesn't get unnecessary access.

What's included in the course

  • 9 video lessons totaling 2 hours and 12 minutes — a recording of a practical workshop split into thematic lessons.
  • A 117-page book — How to Create Loops with Claude Code: a written guide in eight chapters with a table of contents and 34 diagrams.
  • Three ready-to-use projects with loops that can serve as a basis for your own automations.
  • All prompts from the workshop, including /loop commands for scheduling runs and a separate prompt for result verification.
  • Full course presentation — 247 slides on architecture, constraints, triggers, memory, and checks.

Ready projects included

  • minimal-claude-loop — a minimal loop for daily project analysis, created step-by-step during the workshop.
  • Minimal-git-loop — a loop reacting to events in GitHub. It tracks changes in the repository and launches a review only after an actual change in the HEAD branch. Includes a trigger script, state files, and automatically generated reports.
  • Claude-Loop-Test — a test repository used as an observation object for checking the cycle's operation.

What you'll learn

  • Understand when a task really needs a loop and when a single well-crafted prompt suffices.
  • Create a minimal working loop with a clear task, instruction file, state file, and predictable location for results.
  • Organize memory between sessions without turning the state file into an uncontrolled storage of unnecessary information.
  • Separate the execution stage from the verification stage, so the cycle ends not because Claude "considers the task completed," but because the result passed the given validation.
  • Launch loops at specific intervals using /loop and understand the limitations of this approach.
  • Set up event-based launches, such as after a new commit in GitHub, not just on schedule.
  • Integrate external tools and services — GitHub, Slack, Jira — without uncontrolled expansion of AI actions.
  • Design a staged permission system where Claude gets only the access level genuinely required for the current loop.
  • Set stopping conditions, iteration limits, and mandatory human involvement points.
  • Build autonomous AI processes safely: with checks, constraints, and clear accountability.

Practical application scenarios

The course is especially useful for tasks that are regularly repeated and require context preservation between runs. For example:

  • daily or weekly project status analysis;
  • automated review of changes after new commits;
  • preparing technical reports on the repository;
  • triaging tasks, issues, and pull requests;
  • monitoring changes in the codebase;
  • checking pre-defined engineering conditions;
  • preparing structured notes for the team;
  • passing repetitive operations to Claude Code without losing control.

Who this course is for

The course is designed for developers, engineers, and technical specialists who already use Claude Code and want to automate recurring workflows: reviews, project analysis, report preparation, triaging, change verification, and other regular operations.

If you've found yourself thinking: “I run the same prompt every Monday” — this course will show you how to turn such a process into a managed loop with state, checking, and understandable boundaries.

Who the course is not for

The course is not meant for those looking for a no-code tool, a ready-made framework for autonomous AI agents, or a way to hand over development and release changes in production to Claude without human control.

The main principle of the course is the opposite: autonomy should be granted gradually. Initially, a minimal cycle is created, then checks, memory, triggers, integrations are added, and only after that — broader permissions.

What you need to take the course

To take the course, you'll need:

  • installed Claude Code;
  • confident terminal skills;
  • basic knowledge of Git and repositories;
  • a project you wish to monitor or plan to automate.

No additional frameworks, complex infrastructure, or separate API keys other than access to Claude are required.

Course outcome

After completing the course, you'll have not just a set of prompts but an understanding of the architecture of working loops for Claude Code. You'll be able to create cycles that repeat useful actions, retain context, verify results, and operate within safe boundaries.

The main goal of the course is to teach you to build AI automation that can be trusted gradually: from a minimal loop to event-based launches, integrations, and multi-stage result verification.

Who teaches Hands-On Loop Engineering with Claude Course? Youssef Hosni

Youssef Hosni thumbnail

Youssef Hosni — AI Engineer, Data Scientist, and Applied Scientist, a researcher in the field of machine learning and AI. He is currently affiliated with Aalto University in Finland, where he works on the application of AI agents and ML, as well as the development of AI products. On GitHub, he describes himself as “AI Engineer & Applied Scientist | PhD Researcher @AaltoUniversity”.

In his professional description, it is also noted that he works as a Senior Data Scientist at Ment, where he is involved in implementing Generative AI features into the company's products. Previously, he was engaged in research in the field of Deep Learning and Computer Vision, including applications to medical images.

His specialization aligns well with the courses he is currently offering:

  • Data Science and Machine Learning
  • Generative AI and LLM
  • AI Agents
  • RAG and Agentic RAG
  • Prompt Engineering
  • LangChain
  • Claude Code
  • AI Engineering
  • Python
  • Docker
  • Time Series Analysis
  • Computer Vision
  • technical training and preparation for AI/Data Science interviews

What lessons are included in Hands-On Loop Engineering with Claude Course?

This is a demo lesson (10:00 remaining)

Watch the first 10 minutes free. One payment unlocks all 9 lessons here and every other premium course. No auto-renew.

Watch all 9 lessons from $6.67/mo
0:00
/
#1: 01 - Why Loops and the Anatomy of a Claude Loop
All Course Lessons (9)
#Lesson TitleDurationAccess
1
01 - Why Loops and the Anatomy of a Claude Loop Demo
17:31
2
02 - When to Build a Loop and When Not To
17:06
3
03 - The MVP Loop Task, State, and Instructions
13:15
4
04 - Running Your First Loop Manually
15:07
5
05 - Persistent State With progress.md
09:02
6
06 - Verification Separating the Worker From the Checker
18:01
7
07 - Scheduling Loops With the Loop Command
22:52
8
08 - Tools, Connectors, and Safer Permissions
09:41
9
09 - Walkthrough A Real GitHub Repo Review Loop
10:00
Watch the other 8 lessons

One payment unlocks this course and every other premium course. No auto-renew.

See plans

Books

Read Book Hands-On Loop Engineering with Claude Course

#TitleTypeOpen
1How to Create Loops with Claude Code —Unlock
2claude loops workshop —Unlock

What courses are similar to Hands-On Loop Engineering with Claude Course?

More courses by Youssef Hosni

Frequently asked questions

Can I start if I've only used Claude for one-off prompts?
Yes—the course starts with the basics of designing a loop, so you don't need to arrive with an existing automation setup. You'll move from defining a task and its instructions to running a first cycle manually, then adding persistent state and scheduling. The practical work centers on Claude Code and includes a GitHub repository review, so expect an engineering-oriented setting rather than a course focused on everyday chat.
What will I actually practice building?
You'll practice turning a recurring task into a controlled cycle: giving Claude the context it needs, preserving progress between runs, checking its work, and deciding when to continue or involve a person. There's a walkthrough of a GitHub repository review loop to make that concrete. You also get three working loop projects that you can inspect, adapt, and run yourself, rather than having to reconstruct everything from the videos.
Is this useful if I'm already getting good results from my prompts?
It can be, especially if getting those results still means repeating the same setup and checking everything by hand. The focus is less on polishing a single prompt and more on managing what happens across repeated runs. You'll work through questions such as what starts the next cycle, what context carries forward, and how a result gets verified. There's also attention to deciding when a task doesn't need a loop at all.
Does this use Claude Code, or can I follow it entirely in regular chat?
The workflow is built around Claude Code, so you shouldn't expect a chat-only walkthrough. You'll see persistent state handled with progress.md and scheduling through the Loop command. Tools, connectors, and permissions also come into the workflow, with attention to limiting what Claude can change. The GitHub repository review gives you a concrete engineering example, but the course isn't presented as a tool-neutral tour of different AI assistants.
Is this mainly about writing better prompts, or something beyond that?
Prompts are part of the setup, and you receive the workshop prompts, but prompt wording isn't the main focus. You'll spend your attention on the surrounding process: triggers, saved state, verification, permissions, and conditions for stopping or asking for human help. Think of it as learning how to manage recurring work around Claude, rather than a general guide to everything Claude can do or a collection of tricks for individual conversations.
How much time should I set aside, and what comes with the videos?
The videos run for 2 hours and 12 minutes across nine lessons. That's the viewing time; if you want to inspect the projects, try the prompts, or adapt a loop to your own work, allow additional time at your own pace. Alongside the videos, you get a 117-page book, 247 presentation slides, the workshop prompts, and three working loop projects. You can use those resources to revisit the setup while experimenting.
What could I apply afterward beyond the repository review example?
You can carry the same design questions into other recurring engineering tasks: what should trigger the work, what information needs to persist, what changes are allowed, and how will you know the result is acceptable? Separating the worker from the checker is another useful pattern you'll encounter. The broader skill is structuring repeatable AI-assisted work with explicit checks and human handoffs, while the implementation examples remain centered on Claude Code.