Agent School
GitHub ↗

For the Enterprise · Open Source

The hands-on way into
enterprise agentic AI.

Five hands-on courses. Each takes a real Telco-AIX experiment and rebuilds it as a governed, cloud-native agent workload, following the blueprint from our article Architect an open blueprint for cloud-native AI agents: the model is a stateless inference service, the harness owns the loop, and the runtime governs what the agent may touch. Every course is developed and tested on a real enterprise AI platform, captured with narrated walkthroughs, and snapshotted into a self-service interactive lab.

Explore courses ↓ View on GitHub
5Courses
5Narrated walkthroughs
Cloud-nativesafe, dynamic scale, best price/performance
MITOpen source

The curriculum

From a single agent to a governed, self-deciding loop

Each course adds one return-on-investment (ROI) opportunity on the same AI platform investment. The harness grows, the platform does the heavy lifting. Click a course for the problem, the walkthrough video, the architecture & code, and hands-on experience in the interactive lab.

Publications

The thinking behind the build.

Agent School is the working implementation of a body of writing on cloud-native AI agents and distributed inference. Each article below opens on the site where it was published.

Read before you run

Disclaimer: These are enterprise agent workloads, not consumer quickstarts. Each course demonstrates a governed agentic platform: the platform and its domain dependencies (served models, a feature store, a governed tool gateway, workload identity) must be up first. The loop mechanics run offline on a laptop for learning; the real value appears on the platform. Open-source educational example code, provided as-is under MIT, not a supported product. Read the full positioning.