The open-source AI training platform that uses your codebase as the curriculum.
6 academies. 70+ modules. Interactive playgrounds. Free forever.
Enterprise edition: connect your repo, and every example becomes your code, your products, your production incidents.
Every enterprise is paying for two things separately:
- AI/LLM upskilling — generic Coursera courses, external workshops, tutorials disconnected from daily work
- Internal onboarding — learning the codebase, products, systems, and incidents
These are separate budgets, separate teams, separate timelines. Neither works well alone.
One platform that does both. When you teach "how to build a RAG pipeline," don't use a fake e-commerce dataset — use your company's actual knowledge base. When you teach "context engineering anti-patterns," show your own production incidents where context rot caused agent failures.
Your engineers learn AI patterns AND your product domain. One training budget, 2x the value.
Toggle context components on/off and watch the agent's response quality change in real-time. Each playground has 3 scenarios, 6 toggleable components, and quality scores from 0-100.
| Empty (Score: 12) | Partial (Score: 55) | Full (Score: 95) |
|---|---|---|
 |
 |
 |
| Observability Playground | RAG Playground | Multi-Agent Playground |
|---|---|---|
 |
 |
 |
| Evals Playground | Tool Use Playground |
|---|---|
 |
 |
| Academy | Modules | What You Learn |
|---|---|---|
| Context Engineering | 11 | System prompts, tools, memory, RAG, few-shot, conversation history. The 6 core components and 4 strategies (Write, Select, Compress, Isolate). |
| Agent Observability | 9 | Tracing, structured logging, metrics, dashboards, cost tracking. LangSmith, Phoenix, Langfuse, Braintrust. |
| LLM Evals | 13 | Golden datasets, LLM-as-judge, regression testing, CI/CD pipelines. promptfoo, Braintrust, eval-driven development. |
| Agentic RAG | 13 | Chunking strategies, hybrid search, reranking, Self-RAG, corrective RAG, Graph RAG, multimodal RAG. |
| Multi-Agent Orchestration | 12 | Sequential, parallel, hierarchical, swarm patterns. CrewAI, AutoGen, LangGraph. Task decomposition, error recovery. |
| Tool Use & MCP | 12 | Tool design, Model Context Protocol, MCP servers, tool selection & routing, security, production architecture. |
Each academy includes:
- Interactive code examples (bad vs. good patterns with explanations)
- Anti-patterns with severity levels and fixes
- Best practices checklists
- Curated resources (papers, blogs, repos)
- Interactive playground with real-time quality scoring
Your Assets Contextualization Engine Your Academy
─────────────── ──────────────────────── ─────────────
GitHub repos ──┐
Product docs ──┤ ┌──────────────────────┐
Support tickets ──┼───▶ │ Ingest → Chunk → │ ───▶ Domain-specific
Incident reports ──┤ │ Generate examples │ code examples,
API specs ──┤ │ Build scenarios │ real playground
Internal tools ──┘ │ Create anti-patterns │ scenarios from
└──────────────────────┘ your actual systems
| Endpoint | Purpose |
|---|---|
POST /api/ingest |
Accept codebase, docs, tickets, incidents |
POST /api/contextualize/:academy |
Generate enterprise-specific module content |
GET /api/scenarios/:academy |
Return playground scenarios using your data |
GET /api/progress/:team |
Track completion, skill gaps, ROI |
| Feature | Generic Academy | Enterprise Edition |
|---|---|---|
| Code examples | Fake e-commerce app | Your actual codebase |
| Playground scenarios | Generic support bot | Your product workflows |
| Anti-patterns | Textbook examples | Your production incidents |
| Eval datasets | Synthetic queries | Your real user queries |
| Onboarding value | Zero | Engineers learn your product too |
| Layer | Access |
|---|---|
| Open Source Academy | Free forever — 6 academies, 70+ modules, playgrounds |
| Contextualization API | Paid — ingest enterprise assets, generate domain-specific content |
| Hosted Enterprise Instance | SaaS — branded academy with SSO, progress tracking, analytics |
| LMS Integration SDK | Paid — embed into Workday, Cornerstone, or custom LMS |
# Clone
git clone https://github.com/aptsalt/context-engineering-academy.git
cd context-engineering-academy
# Install
npm install
# Dev server
npm run dev
# Build for production
npm run buildOpen http://localhost:3000 in your browser.
- Framework: Next.js 16 (App Router, static export)
- Language: TypeScript (strict mode)
- UI: React 19, Tailwind CSS 4, shadcn/ui
- Icons: Lucide React
- State: React hooks only (no external state library)
- Deployment: GitHub Pages (zero infrastructure cost)
src/
├── app/ # Next.js App Router pages
│ ├── page.tsx # Homepage + enterprise section
│ ├── globals.css # OKLch color system
│ ├── context-engineering-academy/ # Academy page + playground
│ ├── agent-observability-academy/
│ ├── llm-evals-academy/
│ ├── agentic-rag-academy/
│ ├── multi-agent-orchestration-academy/
│ └── tool-use-mcp-academy/
├── components/
│ ├── ui/ # shadcn/ui primitives
│ ├── playground/ # Interactive playground (10 components)
│ ├── chapters/ # Context Engineering modules
│ ├── agent-observability/ # Observability modules
│ ├── llm-evals/ # Evals modules
│ ├── agentic-rag/ # RAG modules
│ ├── multi-agent/ # Multi-Agent modules
│ └── tool-use-mcp/ # Tool Use modules
└── lib/
├── academies.ts # Academy registry (decoupled content layer)
├── data.ts # Content data (swappable per enterprise)
├── playground-data.ts # Playground types & scoring logic
└── scenarios/ # Scenario data (6 academies x 3 scenarios)
All module content lives in TypeScript data files under src/lib/. The content layer is already decoupled from the UI components — enterprise contextualization replaces these data files with domain-specific content while keeping the same UI.
// Generic (open source)
{ title: "RAG Context Selection", bad: { code: "// Dump everything..." } }
// Enterprise (auto-generated from your codebase)
{ title: "RAG for Acme's Knowledge Base", bad: { code: "// How Acme's bot retrieves..." } }PRs welcome. This is an open-source project — contributions to modules, playgrounds, or the enterprise API are all valued.