Reasoning Memory

Reasoning memory is the agent's record of how it works — which tools it reached for, in what order, and how each attempt ended.

In this module, you wrap every turn of your agent in a reasoning trace, make every tool report its steps, and give the agent a tool that consults its own record. The optional lessons record a full trace by hand and read across all three layers with one call.

By the end of this module, you will be able to:

• Wrap every agent turn in a trace that records the task and its outcome
• Make every tool report its steps and tool calls into the open trace
• Give the agent a tool that consults similar past traces
• Record and query a full trace by hand (optional)