Building an AI agent in n8n is easier when you can see each capability added on its own. This pack is a ladder of ten workflows: connect a chat model to an AI Agent, use a Basic LLM Chain and a Summarization Chain, build retrieval-augmented generation with embeddings and an in-memory vector store, add conversation memory, turn on the model web-search tool, then give the agent tools of its own.

The tool workflows show three patterns you will reuse constantly: calling another n8n workflow as a tool (a random-item sub-workflow is included), exposing custom logic through a Python Code Tool, and letting the agent fetch pages with an HTTP Request Tool. Workflow 10 combines all of them with memory in one agent.

All files are plain n8n JSON with credentials, IDs and instance data removed. A start-here guide gives the import order (sub-workflow first), the OpenAI credential you need, and the Data Table and sub-workflow links to re-select. Use the browser to preview and copy any workflow, or download the whole pack as a zip.