AI Agent Memory & Knowledge Kit
8 workflows that fix the three things basic AI agents get wrong: forgetting, hallucinating, and not being able to act
What This Pack Does
Takes an AI agent from a stateless chat loop to one that holds short- and long-term memory, retrieves answers from your own documents, calls real tools, and runs either over Telegram or fully locally.
Who It's For
People who have built a basic AI chat workflow, hit its limits, and need memory, grounded knowledge, and tool use without reading a framework's source code.
Expected Outcome
Ship an AI agent that remembers context, answers from your own documents, and takes real actions
Highlights
- Progression: Foundation → Memory → Knowledge (RAG) → Tools → Deployment
- Short-term and long-term memory, including storage that survives restarts
- RAG wiring so answers come from your documents rather than model guesses
- Tool use via both a direct integration and an MCP server
- A fully local option on Ollama for data that cannot leave your network
Included Workflows(8)
Foundation
Memory
Knowledge
Tools
Deployment
Setup Requirements
An n8n instance and an OpenAI API key cover most of the pack. Calendar tool use needs a Google account, the voice assistant needs a Telegram bot token, and the local deployment workflow needs Ollama running on your own machine.
Example Use Case
A consultancy builds an internal assistant for its team. They start from the agent foundation, add long-term memory so it remembers project context between sessions, and use the RAG workflow to load two years of proposals and delivery notes. The calendar tool-use workflows let it book and reschedule client calls directly. Because some client material cannot go to a third-party API, they run the Ollama variant for that workspace — same architecture, no data leaving the building.
