Durable Execution & Dapr University
Free, hands-on courses to build durable AI agents, workflows and applications based on Dapr
- 11 hands-on courses, beginner to advanced
- Free, and every lab runs in your browser
- 30–50 min per course, in Python, .NET, Java and JavaScript
Choose your learning path
Pick the learning path that matches what you're building. The courses in each path are free and run hands-on in your browser.
Make AI Agents Durable
Keep agents from losing their work when an LLM call, a tool, or the process itself fails.
Courses cover LangGraph, DeepAgents, the Microsoft Agent Framework (MAF) and Dapr Agents, each put on Dapr Workflow's durable execution.
Make Workflows Durable
Write long-running orchestrations that survive restarts, retries and failures.
Task chaining, fan-out/fan-in, monitors, timers, external events and child workflows, in Python, .NET and Java.
Make Applications Durable
Add reliable state, messaging, service calls and secrets to distributed apps with Dapr.
State management, service invocation, pub/sub subscriptions, routing and dead letters, and secret stores, run locally or on Diagrid's managed Dapr.
Collect the University badges
Finish a track and you can claim its digital badge on Holopin. Add the badges to your Holopin board, show them on your GitHub profile, and see how many you can collect. There are 6 to earn so far, so pick a track and start collecting them all.
What you'll learn
Durable AI agents with LangGraph, DeepAgents, MAF and Dapr Agents
An agent framework decides what to do next; none of them survive a crash halfway through. The AI courses take a working agent — a LangGraph graph, a DeepAgents issue investigation, a Microsoft Agent Framework (MAF) multi-agent app in .NET Aspire, a Dapr Agents app with tool calls and MCP — and run it on Dapr Workflow's durable execution. You crash each one mid-run and watch it resume from its last completed step instead of restarting and paying for the same LLM calls twice.
Durable execution and Dapr Workflow patterns
Durable execution means the runtime, not your code, remembers how far a long-running process got. The workflow courses cover it in code you run yourself: activities and retries, task chaining, fan-out/fan-in, the monitor pattern, timers, external events and child workflows, plus the management API for querying, pausing and terminating a run. In Python, .NET and Java, with an advanced course that builds one in .NET with Aspire.
The Dapr building blocks underneath
Durable agents and workflows sit on ordinary distributed-application plumbing. The Dapr courses start with state management, pub/sub and service invocation in .NET, Python, Java or JavaScript. Two deep dives follow: pub/sub subscription types, CloudEvents, bulk messages, routing and dead-letter topics, ending with a swap from Redis to RabbitMQ; and secrets, read through the Dapr Secrets API, scoped per app and moved to HashiCorp Vault without a code change. Another course shows how to run the result on Diagrid without operating Dapr yourself.
What courses should we build next?
Help us prioritize which university courses to build next. Pick your top 3 and we'll let you know when they go live.





