Skip to main content
Maestro is the orchestrator that runs your agent’s conversations. You describe what the agent should do through Skills; Maestro decides which skill is active at each moment, hands off between them, calls your tools, and keeps track of everything it has learned along the way. What makes Maestro different is where control lives. It runs one LLM in a loop, but the framework, not the model, is in charge of that loop. The loop is deterministic-first: Maestro settles everything it can on its own before calling the model, and checks every model decision against the rules you’ve set. That’s what lets you start with plain-language instructions and add framework-enforced guarantees only where a mistake would be expensive, without rewriting the skill. In one line: Maestro turns a set of skills into a coherent, controllable conversation, with the framework holding the guarantees and the LLM handling the judgment.

​
A deeper look

The guides below take that apart, one piece at a time.

The Runtime Loop

What happens on every turn, step by step, and why the framework runs first.

Guarantees & Guardrails

What the framework enforces itself versus what it leaves to the LLM’s judgment.

Context Management

How Maestro stays coherent across a long conversation without overwhelming the model.

Skills & Routing

How Maestro picks a skill, stacks sub-skills, and handles a user who changes topic.