Cycode Maestro is evolving from an AI orchestrator into your personal agentic security companion.
A tool starts fresh with every task. A companion remembers what you’ve done together, learns how you work, and builds on that experience over time.
Today, we’re introducing personalized memory for Maestro, the capability that makes this transformation possible.
In one of our previous posts, we explored how security teams use Maestro, the questions they ask, and the work they delegate to it.
Through the Cycode platform, Maestro already has access to your company’s software development assets. From them, it infers how your software is structured, how it runs, and which assets matter most to protect. But the unwritten context security teams share during investigations like tribal knowledge about legacy systems or business blast radius often didn’t carry over.
For example, during an investigation into payment services, an engineer explained that an unmaintained legacy processor still quietly handles live refunds in production. That context completely changed the risk severity of any credential exposed in that codebase. But without memory, that architectural reality had to be re-explained every time a new ticket came up.
With personalized memory, that context sticks:
Engineer: “Audit payment-service for hardcoded credentials.”
Maestro: “During our last investigation, you mentioned that the legacy payment processor still handles refunds in production. I’ll check whether any of these secrets could allow unauthorized refunds.”
How it works
We gave Maestro 2 new memory stores: always-on memory and recalled memory.
- Always-on memory holds behavioral knowledge. It learns your preferences, your rules, and even your team’s vocabulary. Maestro applies it in every conversation.
- Recalled memory holds factual knowledge. It learns what matters for each product, what happened in past conversations, and even the context that your tools cannot see. Maestro brings it back when it is relevant.
Together with Maestro’s existing capabilities, these stores allow it to follow the current conversation, understand your environment, adapt to how you work, and remember what happened before. These capabilities map to the four types of memory described in cognitive science:
| Memory type | In cognitive science | In Maestro |
|---|---|---|
| Working | What you hold in mind now | The current conversation |
| Semantic | What you know | Recalled memory: what matters for each product, and the context that your tools cannot see |
| Procedural | How you do things | Always-on memory: your preferences, your rules, and your team’s vocabulary |
| Episodic | What happened, and when | Recalled memory: what happened in past conversations, and when |
You’re in control
Memory should make Maestro more useful, without taking control away from you. Maestro shows you when it saves a new memory, and you can ask it to search, modify, or delete what it remembers. These requests are handled through the same conversation, subject to our safety guardrails. You decide what Maestro should remember and what it should forget.
Agentic security that understands you
Security tools are good at telling you what’s wrong. But anyone who has run a security program knows there’s no single recipe for getting security right.
Every organization has different priorities, constraints, and ways of working. And every security practitioner brings their own experience and judgment to the table.
We don’t want Maestro to replace that judgment. We want it to amplify it. That’s what personalized memory makes possible.
Schedule a demo to see Maestro in action.