An agent with no memory forgets you between sentences. Today you give it a memory and a clean set of tools, and learn why a tool’s description matters more than its code. You’ll leave saying: “It remembered what I said and reached for the right hand.”
Begin →Give the agent short-term (this chat) and long-term (across chats) memory.
Describe a tool so well the model always knows when to call it.
Complete a task that needs two tools and one remembered fact.
Open each part to see what it does and the one mistake beginners make. This is the whole architecture — memorize it and you can read any agent framework.
The model decides whether to call a tool from its description alone — it never reads your code. Pick the description an agent can actually route to.
You have a tool that converts currencies. Which description works best?
Give it hands and a memory, not just a mouth — and describe those hands clearly, because the description is the interface.
You gave an agent working memory and multiple tools, and kept it coherent across a multi-turn task.