Skip to main content
Ultron doesn’t reset between sessions. Every research output, every confirmed preference, every error and its fix — all of it is stored, organized, and recalled automatically. Memory is what separates Ultron from a stateless chatbot. Agents get smarter with every run. Skills produce better output over time. The more you use Ultron, the more it knows about your business.

What gets stored

Memory entries are plain markdown files. Each is tagged with type, agent, and topic. The index (MEMORY.md) gives Ultron a fast overview of everything stored.

Four memory types

Feedback memory stores both errors and validations. When you confirm something works, Ultron stores that too. This prevents behavior drift — Ultron doesn’t unlearn what’s working.

How memory is used

At the start of every agent turn, a separate Sonnet-powered side-query runs against the memory store. It selects the 5 most relevant entries for the current task. Those entries are injected into context before Kimi K2 sees the message. You don’t manage this. It happens automatically.

Memory Layers

The four memory types and when each is created.

Brain Graph

Visual map of everything Ultron knows about your business.

Memory Recall

How Ultron selects the right memories per turn.

Compression Engine

Three-layer cascade that keeps sessions running indefinitely.