P01Runnable
AI Provider Gateway
Provider abstraction, streaming, structured outputs, and trace-aware model access.
TypeScript, Express, Zod
AI engineering operating system6 runnable systemsCI validated
Build production-gradeAI agents, copilots,and tool-using systems.
Agentic Engineering Playbook teaches real AI agent engineering through six runnable projects covering LLM gateways, RAG, orchestration, MCP-style tools, operator UI, and browser agents.
LLM GatewayRAGOrchestrationMCPAgentic UIBrowser Agents
Agent runtime consoleLive pipeline
1
User request
Generate a secure travel copilot for finance.
2
Planner
Break work into gateway, RAG, approval, and QA steps.
3
Retrieval
Pull policy context and prior incident examples.
4
Tool call
Fetch internal docs and repo metadata through safe tools.
5
Approval
Pause on sensitive actions and capture operator reason.
6
Eval
Check grounding, latency, and execution quality before release.
run agent --mode orchestrate --env testFinal answer
Production path ready
Gateway selected, RAG citations attached, approvals enforced, and QA run staged for release confidence.
Latency 842msTools 3Eval pass
Completed projects
Six runnable systems, each teaching a different layer of agent engineering.
This is a buildable curriculum, not a notes dump. Every project is scoped to teach one production-shaped system with local validation and a clear upgrade path.
P02Runnable
Enterprise RAG Copilot
Chunking, retrieval, citations, grounded answers, and eval-first search quality.
TypeScript, Express, JSON storage
P03Runnable
Agent Workflow Orchestrator
State machines, approvals, retries, tool routing, and inspectable traces.
TypeScript, Express, workflow runtime
P04Runnable
MCP Enterprise Toolkit
Safe read-only tools, schemas, resources, and audit logging for enterprise agents.
TypeScript, Express, MCP-style APIs
P05Runnable
Angular Agentic Copilot
Operator UX, session state, streaming UI, approvals, and visible agent activity.
Angular, TypeScript, local mock runtime
P06Runnable
QA Browser Agent
Safe browser automation, evidence capture, dry-run policy, and deterministic QA evals.
TypeScript, Playwright, Express
Learning path
Start with runtime basics, then compound them into full agent systems.
The path is intentionally cumulative: each project becomes infrastructure for the next layer of agent capability.
1
Layer 1LLM Gateway
2
Layer 2RAG
3
Layer 3Orchestration
4
Layer 4MCP Tools
5
Layer 5Angular Copilot
6
Layer 6QA Browser Agent
Why this is different
A serious AI engineering academy, not a generic documentation site.
Build-first, not notes-first
Every layer is tied to a runnable system so learners move from reading to implementation quickly.
Production-shaped from day one
The projects emphasize boundaries, validation, traces, and upgrade paths instead of toy demos.
Evals and CI included
Projects ship with smoke checks, eval flows, or build validation so quality is part of the curriculum.
Agent UX is visible
Plans, approvals, tool activity, and browser evidence are surfaced as operator-facing product behavior.
Safety and approvals are core
Read-only tools, dry-run policy, approvals, and environment controls are treated as first-class patterns.
Portfolio-ready projects
The sequence is designed to produce public work samples that show system thinking, not just prompt experiments.
Architecture preview
See how the systems connect into one agent engineering stack.
UI, orchestration, retrieval, tools, browser automation, and model access are taught as connected layers instead of isolated demos.
Captured demo assets
Real local screenshots now back the launch story.
These screenshots come from the local production docs build, the Angular copilot UI, and the QA Browser Agent report flow. Walkthrough GIFs are still a follow-up asset.
P05 captureAngular Copilot UI
P06 captureQA Browser report view
homepage_smoke
Dry run generated a deterministic QA plan for mock://homepage without launching a live browser session.
Checkdry_run_plan_created
Trace stepsrequest_accepted → dry_run_only
Safety modeNo browser launched
System viewProject architecture preview
Community growth
Make the academy more useful for contributors, learners, and public builders.
The repo is designed to support open-source contribution, practical chapter writing, weekly shipping momentum, and portfolio-quality project submissions.
Contribute a chapter
Propose a new curriculum page with examples, diagrams, exercises, and project tie-ins.
Open the chapter guideTake the weekly challenge
Ship one practical improvement per week and turn the repo into a consistent proof-of-work engine.
See the challenge formatSubmit a project idea
Pitch a runnable project that extends the academy into new agent, UI, eval, or platform territory.
Use the submission templateRecruiter value
This academy is structured to create visible hiring signal.
It helps developers show architecture judgment, implementation range, and production awareness across multiple AI system layers.
Evidence of agent architecture thinking, not just prompt experimentation.
Runnable projects covering copilots, RAG, MCP, observability, and UI workflows.
Public documentation that shows tradeoff thinking and collaboration readiness.
Start building