Scouts: eyes, ears, and a voice in the world

Most of what a flow deals with is data and documents. A Scout is how a flow reaches past that, into the physical world — a loading bay, a shelf, a reception desk, a site gate.

A Scout today is a phone or tablet with the Scout app installed and paired to your workspace. It’s not a robot — nobody drives it, and it doesn’t move on its own — but it gives a flow the four things a person standing there would have: a camera, a microphone, a speaker, and a volume knob. The same model is built to grow into fixed sensors and, eventually, machines that do move and act — see Levels: what kind of Scout is this below.

What a Scout can do today

Capability	What it does	The flow step
Capture	Takes a photo	Scout Capture
Listen	Records a short audio clip	Scout Listen
Speak	Says something out loud	Scout Speak
Set volume	Changes how loud the device is	Scout Set Volume

Each one is both a step you drop into a flow and a tool an agent can call — the exact same command either way, so a device behaves identically whether a human built the automation or an agent is driving it.

There’s a fifth thing a Scout with both Speak and Listen can do: hold a two-way conversation. That’s not a module of its own — it’s the Ask a person step’s Voice (Scout) channel, which speaks a question and listens for the spoken reply in one step. See Asking a person for how it works.

The one thing to understand: it’s a command, not a phone call

A Scout is not always in reach — it might be face-down on a desk, out of signal, or its screen off. So talking to one isn’t a function call that returns an answer; it’s a command you enqueue, which the device picks up and acts on whenever it next checks in.

That split matters differently depending on what you’re asking for:

• Capture and Listen are reads — you want something back (a photo, a clip). The step enqueues the request and returns immediately with enqueued: true and a command id, but the photo or clip is not part of that step’s output. The device takes its time, uploads the result, and that upload fires a scout.observed event. Your flow reads the result by adding a Wait step after the capture/listen step with event: scout.observed, then referencing {{ wait_event_payload.memory_item_id }} downstream.
• Speak and Set volume are fire-and-forget — there’s nothing to read back, so the step completes as soon as the command is queued. If you do want confirmation that the device actually said it or changed its volume, add a Wait on scout.spoken (after Speak) or scout.command_resulted (after Set volume).

The shape to remember:

Scout Capture (or Scout Listen)  →  Wait on "scout.observed"  →  steps that use the photo/clip

See Wait for the step itself, and Anatomy of a flow for how triggers, steps, and events fit together generally.

What happens to a photo or clip

A captured frame doesn’t just disappear once it lands — it’s stored the same way any file you upload is, with the same automatic OCR and metadata extraction. But it isn’t automatically kept forever: by default it’s temporary evidence that expires after a few days (your workspace’s default, changeable per-capture up to 14 days) — the point is that a Scout can check something constantly without slowly filling up your Memory with routine snapshots. Ask for keep: memory on the step (or click Keep in Memory on a photo later) when a capture is actually worth keeping — a scanned document, proof of delivery — and it becomes a permanent record with full extraction, right away.

Where you work with Scouts

• Settings → Scouts is where a device is paired, renamed, re-placed, or revoked, and where fleet-wide defaults (default camera, resolution, flash, pairing level, token expiry) live. See Pair and use a Scout device.
• Scouts, in the Automations part of the sidebar, is the live monitoring window: every paired device, what it’s doing right now, and — for whoever holds Manager, Approver, or Administrator access — buttons to act on it directly (take a photo, make it talk, record sound, change its volume) outside of any flow. Same page as the guide above covers.

Levels: what kind of Scout is this

Routario groups devices into a level, from what they’re allowed to do:

• L1 · Fixed — senses only (capture, listen, and reserved sensing scopes not yet shipped as steps).
• L2 · Interactive — adds speak, set volume, and reserved local-actuator scopes (display, signal).
• L3 · Mobile and L4 · Manipulating — reserved for devices that move or physically act (locomote, actuate) — not yet available as flow steps.
• L5 · Humanoid — a form-factor label an operator sets by hand, not something Routario derives.

Today, the four capabilities above (Capture, Listen, Speak, Set volume) are what actually run. The higher levels describe where the platform is headed, and the same pairing and Wait-on-event model will carry over once they ship.

Access

Like everyone and everything else in Routario, what a Scout is allowed to do is a grant, not a default — a device only holds the specific capabilities it was paired with, and a flow step aimed at a capability a device doesn’t have simply reports “no device found” rather than failing. See Access: who can do what.

Where to go next

• Pair and use a Scout device — the step-by-step: pairing, the live monitor page, and wiring a Scout into a flow.
• Scout Capture, Scout Speak, Scout Listen, Scout Set Volume — the full field reference for each step.
• Asking a person — the Voice (Scout) channel, for a two-way spoken exchange.
• Wait — the step that resumes a flow when a Scout’s result lands.
• Scout L0 Gate (coming soon) — a different shape of Scout: a small program for a computer already on your network, for reaching a shared drive or a fixed camera rather than carrying a device around.