Cortex
Cortex is a public working library for learning how to build with AI agents.
It starts from a practical question: what changes when AI is allowed to work with the information, tools, and procedures that make up a real project? The answer is not a magical “agent.” It is a system: a knowledge base the agent can consult, useful connections to the outside world, clear instructions for recurring work, and a person who remains responsible for judgment.
This site is the course material for that system. It is written to be read, questioned, and used. The notes are deliberately small. Each one should give you an idea you can test in your own work; the links show how the ideas fit together.
How to use this site
If you are new to the subject, begin with Start Here. Then follow the units in order. If you already have a question—how an agent remembers, how it uses a calendar, or how you keep it from quietly making a mess—use search and follow the links around that question.
The course
- Memory & Knowledge Graphs — build a durable source of context that both people and agents can use.
- Connections — give an agent carefully bounded access to useful tools and information.
- Skills & Personas — turn a repeatable way of working into instructions an agent can follow.
- Autonomy — decide what may run without a prompt and how its work gets checked.
- Leverage in Practice — put the system to work on real opportunities without handing over the judgment that gives the work direction.
What you should leave with
By the end, you should be able to explain the difference between a chat session and an agentic system; create a small knowledge base for a real project; connect an agent to it safely; and show someone else what you built.
Resources collects the public tools and references used along the way.