Configure an agent

An agent is shaped by a handful of settings. The exact options are expanding during beta, but every agent is built from the same basic levers.

What it does

Every agent runs on instructions — plain-language directions for what to watch for and what to do about it. The instructions are the core of the agent: change them, and you change its job.

When it runs

An agent runs on a schedule — daily, hourly, or a custom cadence — or on an event trigger: a notification, a webhook, or a file upload. Some agents instead run manually, firing only when you ask.

What it can do

An agent is given a set of tools it's allowed to use. For each tool, a tool policy decides how much freedom the agent has: some actions it can take on its own, and others need your approval before it acts. That balance is yours to set, so the agent's autonomy stays where you want it and you stay in the loop on anything higher-stakes.

Scope

Agents fall into two groups. Nash agents are built and maintained by Nash out of the box. Personal agents are ones your organization builds and owns.

Where results go

Every run feeds into the Daily Brief — what ran, what it found, and what's awaiting your decision. An agent can also send its results straight to a destination you point it at, like a channel or an inbox, and you can choose to be notified on every run or only when one fails.

Options here vary by organization and plan while Nash Agent is in beta.

See what Nash Agent is for how agents fit into your operations overall.