AI Agents & LangChain in n8n

AI Agents · 12 lessons

Build production AI agents: memory, tools, RAG, structured output, guardrails, evals and multi-agent orchestration.

Start lesson 1
  1. 1AI Agent node basicsChat model + tools + memory: the three building blocks of every n8n agent.40 min
  2. 2Chat Trigger + conversational memoryBuffer, window, and Postgres-backed memory strategies for multi-turn chat.35 min
  3. 3Tools as sub-workflowsExpose any n8n workflow as a callable tool the agent can invoke on demand.35 min
  4. 4Vector storesPinecone, Qdrant and Supabase pgvector compared, with insert and query patterns.40 min
  5. 5Building RAG pipelinesIngest → chunk → embed → store → query → cite: the full retrieval-augmented generation loop.50 min
  6. 6Choosing a model: OpenAI, Anthropic, Gemini, localTrade-offs between hosted frontier models and self-hosted open models via Ollama.25 min
  7. 7Structured output & parsersForce the model to return valid JSON matching a schema you control downstream.30 min
  8. 8Guardrails, evals & testing agentsRegex filters, LLM-as-judge scoring, and repeatable eval workflows before shipping an agent.40 min
  9. 9Human-in-the-loop approvalsPause an agent for a Slack or email approval before it takes a risky action.30 min
  10. 10Cost & latency controlCache repeat calls, pick cheaper models for easy steps, and batch expensive LLM calls.25 min
  11. 11Multi-agent orchestrationRoute requests between a planner agent and specialist sub-agents for complex tasks.45 min
  12. 12AI capstone: support-ticket triage botA production-grade agent that classifies, answers or escalates support tickets with RAG and guardrails.90 min
Next track: Self-Hosting & DevOps