Learning Path
Four progressive stages — each stage unlocks the next. The QA pipeline example grows with you from a single node to a full production system.
Stage 1
Foundations
Modules 1–2
Modules 1–2
Stage 2
Routing & Tools
Modules 3–4
Modules 3–4
Stage 3
Multi-Agent
Modules 5–7
Modules 5–7
Stage 4
Production
Modules 8–10
Modules 8–10
Course Modules
Each module is a self-contained interactive page with concept notes, code examples, flashcards, MCQs, and Q&A — all wired to the QA pipeline running example.
00
New to LangChain or async Python? This 30-minute bridge covers the exact four LangChain primitives used in the course — nothing more. If you've called an LLM API before, you can skim or skip.
async / await
TypedDict
ChatAnthropic
Message types
@tool
bind_tools
01
Everything you need to understand LangGraph's mental model — how it differs from LangChain, why state machines matter for agents, and how to build, compile and run your first graph.
StateGraphNodes & Edges
TypedDict stateSTART / END
compile & invokeMermaid viz
02
Design state schemas that scale. Learn reducers, message history with add_messages, and how MemorySaver and SqliteSaver give your pipeline persistent memory across CI runs.
Reducersadd_messages
MemorySaverSqliteSaver
thread_idResume from checkpoint
03
Move beyond linear graphs. Build router functions that branch on state, create fan-out / fan-in patterns, and route the QA pipeline based on test failure severity.
add_conditional_edgesRouter functions
Fan-out / fan-inLiteral types
Short-circuit on failure
04
Bind tools to LLMs, use ToolNode, and wire your existing MCP servers into a LangGraph workflow. The QA example connects knowledge-store-mcp and playwright-mcp as tool nodes.
bind_toolsToolNode
ReAct in LangGraphMCP → LangGraph
Error handlingTool selection
05
Add approval gates before destructive actions. Use interrupt() to pause execution, edit state mid-graph, and resume — essential before auto-creating Jira tickets or posting Slack alerts.
interrupt()Breakpoints
State editingCommand.RESUME
Approval workflows
06
Design supervisor hierarchies, implement agent handoffs, and share state across specialist agents. The CI/CD Orchestrator becomes a proper LangGraph supervisor coordinating all 14 agents.
Supervisor patternAgent handoffs
Shared stateHierarchical agents
Command objects
07
Replace asyncio.gather with LangGraph's Send API for proper map-reduce. Encapsulate each pipeline phase as a subgraph with namespaced state — making the architecture inspectable and resumable.
Send APIMap-reduce
SubgraphsState namespacing
Result aggregation
08
Trace every agent decision in LangSmith, set up dataset-based evaluations, debug failed runs at the node level, and build cost and latency dashboards for the CI pipeline.
LangSmith tracingRun trees
Datasets & evalsCost tracking
Feedback logging
09
Close the evaluation gap. Build a DeepEval metric suite for agent decision quality, use RAGAS for RAG retrieval quality, and embed evaluation as a first-class LangGraph node in the CI pipeline.
DeepEval metricsRAGAS
FaithfulnessAnswer relevancy
Eval as graph node
10
Rebuild the complete 3-phase CI/CD QA pipeline as a LangGraph StateGraph — MCP servers as tool nodes, HITL approval gates, LangSmith tracing, and a DeepEval evaluation layer. The final production system.
Full StateGraph7 MCP tool nodes
14 agentsHITL gates
LangSmithDeepEval
What You'll Build
By Module 10 you will have a fully working LangGraph-native version of the K11 TechLab QA automation system, rebuilt with proper state management, HITL gates, and observability.
3-Phase CI/CD QA Pipeline
A LangGraph StateGraph where each phase is a typed subgraph, agents are nodes, MCP servers are tool layers, and LangSmith traces every decision.
1
Planning — SequentialTest Blueprint Agent → Test Generation Agent
2
Execution — Parallel (Send API)E2E · API · Perf · Visual · A11y · Security · Data Guard
3
Analysis — SequentialDefect Analyst → CI Monitor → Coverage → Report Gen
All 14 Agents
Test BlueprintTest Gen
E2E RunnerAPI Probe
Perf MonitorVisual AI
AccessibilitySecurity Scan
Data GuardDefect Analyst
Chatbot EvalCoverage
CI OrchestratorReport Gen
7 MCP Tool Nodes
knowledge-store
playwright
k6-load
github
jira
postgres
slack
Prerequisites
You don't need prior LangGraph experience — but you'll get more from the course if you have these foundations.
Python 3.11+
Comfortable with async/await, dataclasses, type hints, and virtual environments.
LLM API basics
Understand how to call Claude or OpenAI — messages, roles, tool use / function calling.
Agent concepts
Know what a ReAct loop is. Module 1 recaps it, but prior exposure helps.
MCP (optional)
Module 4 covers MCP integration — useful if you've seen the K11 TechLab QA platform.
Enroll in This Course
One-time payment. Lifetime access. All future module updates included.
What's included
10 interactive modules (80+ code examples)
Flashcards, MCQs, and Q&A for every module
Full capstone project — 14-agent CI/CD QA pipeline
LangSmith & DeepEval integration walkthroughs
Access to the live system demo
All future content updates at no extra cost
Certificate of completion