Meet Nova: the AI manager that orchestrates your workforce

Every AI workforce needs a manager. Someone who knows the whole operation — which agents exist, what they are capable of, what data they can access, what work is in progress — and can translate a human request into coordinated action across the entire team. That is Nova.

Nova is not another agent in your workspace. It is the layer above them. The orchestration intelligence that sits between what you want done and the workforce that does it.

It already knows your workspace before you say anything

When you open a conversation with Nova, it has already assembled a complete picture of your environment. Every office in your workspace and what each agent is configured to do. Every data namespace and what it contains. Every integration that has been connected. Every cron job scheduled. The current state of ongoing work. Nova does not ask clarifying questions about your setup because it already knows your setup.

This context is live — not cached from yesterday. If you added an agent this morning, Nova knows about it. If an integration was connected an hour ago, Nova can already work with it. The workspace state is what Nova sees, and it is always current.

That awareness is what separates an orchestration layer from a chat interface. A chat interface responds to what you say. Nova responds to what you say in the context of everything it already knows. The difference in practice is enormous. You do not need to explain your setup every time you want something done.

From a sentence to coordinated action

When you give Nova a task, it does not just answer — it acts. It determines which agent or agents are best suited to handle the work, wakes them if they are sleeping, dispatches the right instructions, and tracks progress through to completion. You get a result, not a plan for how you could get a result.

Complex work gets decomposed. If you ask Nova to research a competitor, write a summary, and store it in your data plane, it breaks that into steps, routes each step to the appropriate agent, and sequences the work. The agents operate in parallel where they can. Nova holds the thread across all of them. When the work is done, Nova reports back with what happened.

Every action Nova takes is visible. You can see which agents it dispatched work to, what it told them, and what came back. Nothing happens silently. If a step fails, Nova surfaces it. If something requires a decision before proceeding, Nova asks rather than guessing.

Guardrails without friction

Autonomous systems that act without sufficient caution are dangerous. Nova is designed to be genuinely useful without being reckless.

Destructive actions — deleting data, modifying configurations, taking irreversible steps — require explicit confirmation before Nova proceeds. This is not a blanket approval gate on everything. Nova uses judgment about what constitutes a high-stakes action and surfaces those specifically. Low-stakes work flows without interruption. High-stakes work pauses for human sign-off.

When an agent needs credentials it does not have — an API key, a service token, access to a system — Nova surfaces a structured request with context explaining what is needed and why. Not an error. Not a log message. A form the user fills out, with the credentials saved directly to the vault. The agent then has what it needs and continues. The human provided the input without ever having to understand the underlying infrastructure.

Voice as a first-class interface

Nova supports voice natively. Speak a request, get a spoken response. The voice pipeline handles transcription, routes the input through the same agent loop as text, and converts the response back to audio. The agent does not know or care whether the input was typed or spoken — the orchestration layer is identical either way.

This matters for users who are not sitting at a keyboard. Reviewing work on a phone, checking in while commuting, giving a quick instruction between meetings. Voice makes the workforce accessible from anywhere, not just from a laptop with a chat window open.

Personality that fits your organization

Nova can be configured with a tone and set of behavioral instructions specific to your workspace. If your team communicates formally, Nova communicates formally. If you prefer directness over explanation, Nova adjusts. Custom instructions let you shape how Nova presents information, what it prioritizes, and how it handles ambiguity.

This is not cosmetic. The way an orchestration layer communicates determines whether the people using it trust it. A Nova that sounds like it belongs in your organization gets used. One that feels generic or off-tone gets worked around.

From beginner to expert, same interface

Nova is designed to work for people at very different levels of familiarity with AI systems. A person who has never used an agent platform before can describe what they want in plain language and get useful work done. An engineer who understands the full stack can give precise, technical instructions and get exactly what they specified.

The interface does not require you to know how the workforce is organized before you can use it. Nova handles the routing, the waking, the sequencing, and the reporting. You describe outcomes. The infrastructure handles execution.

That is the idea behind Nova: make the AI workforce as easy to direct as telling a colleague what you need done.