Applied Agentic Software Engineering
EECS 498-016
OpenClaw is an open-source platform that connects chat apps to AI agents that code, research, and coordinate autonomously — agent loops, tool registries, session routing, multi-channel message gateways. This course teaches you to build systems like it, from scratch.
You won't just use AI tools — you'll understand the fundamental patterns behind production agent platforms. By the end, you'll have designed, implemented, and defended your own orchestrator using the same principles that power systems like OpenClaw.
No toy demos. No prompt-engineering-only courses. Real systems, real code, real engineering.
4 Credits
Full technical elective course with hands-on projects
Prerequisites
EECS 281, EECS 201 or ULCS or Instructor Permission
Schedule
Tue/Thu lectures + weekly lab sections
Format
No exams—all projects, demos, and oral defenses
Inspired by a Real Platform
Most courses teach agent concepts in a vacuum. This one is grounded in the architecture of a real, production open-source system — so every pattern you learn maps directly to how agent platforms actually work.
Production Reference
OpenClaw is a real open-source multi-agent gateway — session management, tool execution, plugin architecture, multi-channel routing. It's the reference architecture that grounds every concept in this course.
Same Patterns, Your Code
You'll build your own implementations of the core concepts — agent loops, tool registries, delegation, orchestration — learning the same patterns that power production platforms.
Ready to Contribute or Create
By the end, you'll deeply understand how platforms like OpenClaw are architected. You could contribute to OpenClaw, extend it, or build the next one.
Three Phases, One Goal
Each phase builds on the last, taking you from AI-assisted coding fundamentals to building a complete agent orchestration platform — learning the same patterns that power production systems like OpenClaw.
AI-Assisted Development
Master the fundamentals of AI-assisted coding through three core principles: Context, Model, and Prompt. Learn to use AI tools with the same principled context management and prompt engineering that powers agents in platforms like OpenClaw.
- 6 lectures + 3 labs
- Individual project demonstrating principled AI coding
- Build real software with AI assistance using Aider
Scripted Agentic Workflows
Build multi-agent communication protocols from scratch — the same patterns OpenClaw implements for session routing and message gateway coordination. Design isolated state, structured handoff protocols, and pluggable LLM backends.
- 6 lectures + 3 labs
- 4 incremental milestones with checkoffs
- Live multi-agent system demonstration
Build an Agent Orchestrator
The capstone. Design and implement a production-grade orchestrator using the same architectural patterns as OpenClaw — task systems, tool registries, agent delegation, and evaluation. Build five core components, then extend with 8+ capability milestones chosen from 40+ options. Defend your work in an oral exam.
- 16 lectures + 8 labs
- Modular system with evaluation report
- Oral defense demonstrating mastery
What You'll Build
This isn't a theory course. Every week you'll ship working code. Here's what you'll have by the end:
AI Coding Portfolio
A collection of projects demonstrating effective AI-assisted development. You'll use Aider to build real software, document your process, and develop repeatable techniques.
Multi-Agent Orchestrator
A working system that coordinates multiple AI agents with file-based state management, structured protocols, and pluggable LLM backends — the same architecture pattern used in platforms like OpenClaw.
Modular Agent Platform
Your capstone: a complete agent orchestration platform with task management, tool execution, delegation, evaluation, and 8+ custom capabilities — built from scratch using production-proven patterns.
Evaluation Report
A technical report analyzing your system's performance with empirical data, comparative benchmarks, and thoughtful analysis of design tradeoffs.
Oral Defense
Present and defend your final project to faculty. Demonstrate technical mastery, explain design decisions, and answer challenging questions about your implementation.
Deep Understanding
Beyond code: a mental model for agentic systems. You'll understand how agents coordinate, when to delegate, how to evaluate, and what makes systems production-ready.
Instructors
Learn from instructors at the forefront of AI and software engineering.
Marcus Darden
Instructor
mmdarden@umich.edu
Rafe Symonds
GSI
rsymonds@umich.edu