Useful things, not a catalogue
Resources worth opening.
A small set of tools and primary sources. Free tiers, product names and protocol details move quickly; when a decision matters, use the source link and check the date.
Build and inspect
The small set of tools used by the Workbench and lessons.
Python
The language used for the Workbench. Start with standard-library code before frameworks.
Streamlit
A thin local browser view for evidence, state and decisions. It is not a Docker requirement.
OpenCode
A coding companion for setup, small edits, review and learning from a working repository.
Groq
The planned default model connection once the course reaches model-assisted behaviour. Free-tier limits change, so check the current account page.
Specifications and primary references
Read the source when the implementation detail matters.
Model Context Protocol
The specification and reference implementation for model-context interoperability.
Agent2Agent Protocol
The protocol for communication between independent agent systems.
OpenTelemetry
Specifications for telemetry signals and semantic conventions.
OpAMP
Remote management for fleets of telemetry data-collection agents. These are not LLM agents.
Reliability and security
The foundations we will carry into the production path.
Google SRE book
The Four Golden Signals and the reasoning behind useful monitoring.
OWASP Agentic Top 10
A current threat model for agentic applications.
NIST AI RMF
A framework for managing AI risks across a system lifecycle.
Kubernetes security
Official guidance on RBAC, policies, secrets and workload controls.