Skip to main content

The Operator

Coming soon. The Operator is not available yet. This page describes what it is for and how it works, so you can plan for it; it will say how to get the Operator once it is released.

The Operator is Floxar's ready-made agent for running your processes. Where the AI Assistant answers questions and the Flow Designer drafts flows, the Operator runs them: it takes in a piece of work, finds the flow your team wrote for it, works through the steps as a trail, and either finishes the run or hands it to a person with everything they need to take over.

It is meant for teams that want AI agents to carry part of their workload but do not already run agents of their own. If you already have an agent that works with Floxar, you do not need the Operator; it is one way to run flows with an agent, not the only one.

How a run starts​

The Operator starts work in one of four ways:

  • On request. Someone, or another system, asks it to run a specific flow.
  • From a queue. It watches a team queue and picks up trails placed there, so your team can hand it work the same way they hand work to each other: by putting a trail in the queue or transferring it to the Operator.
  • From an event in another system. A system you use notifies it that something happened, such as a new ticket or form submission. The same event delivered twice never starts two runs.
  • On a schedule. It runs a flow at set times.

How it chooses a flow​

Either the request names the flow to run, or the Operator looks through your account's flows and picks the one that fits the work. It only ever runs flows your team has written; it never makes up a process. When the choice is clear, it starts the run and records on the trail why it chose that flow. When several flows could fit, it hands the decision to a person, with the candidates. When nothing fits, it reports the work as unmatched so your editors can see what kind of work has no flow yet, and runs nothing.

How it works through a flow​

The Operator works the way anyone running a trail does, one step at a time:

  • It reads the current step, does what the step describes, and records its answers on the trail like any person would. At a decision point it chooses among the paths your authors defined and notes why.
  • Before every action it checks the trail again, because a person may have acted on it in the meantime.
  • It never invents an answer to get past a required field. If it cannot give a real answer, it hands the trail to a person.
  • When a step needs to act on one of your systems, it uses credentials from the Secrets Vault, only for that moment and only if it has been given access to that particular credential. Credentials never appear on the trail. If a credential is missing, it hands off instead of working around it.
  • An action in another system that cannot safely be repeated or checked afterwards is left to a person by default.
  • Limits on the number of steps, repeated steps and time stop a run that is going nowhere and hand it to a person.

Handing work to people​

For the Operator, handing a run to a person is a successful outcome, not a failure. Whenever it reaches something it should not decide alone, it:

  1. Writes a hand-off note on the trail: what it tried, what it knows, why it stopped and what the person needs to decide.
  2. Transfers the trail, to your team's queue by default or to a named person.

Whoever picks the trail up sees the full record of what the Operator did, step by step, and continues from where it stopped. They can finish the run themselves or, when the decision is made, put the trail back in the queue for the Operator to continue. See Running and tracking trails.

When something goes wrong​

If the Operator is interrupted partway through a run, it checks the trail to see what actually happened before continuing, and carries on from the step the trail is really at. If it cannot tell whether an action went through, it hands the trail to a person rather than risk doing it twice.

How it fits with the rest of Floxar​

The Operator is a registered agent of your account: it has its own identity and role, its work is shown under its own name marked as an agent, and it follows the same permissions, request budget and capacity as any other agent. It runs flows but never writes or changes them. Your account's administrators register it like any other agent; see Registering an agent and Agents and AI clients.

Last reviewed: 2026-10-06