Intelligence
Scout
Scout is a supervised diagnostic session on a customer's Windows device. You give it an objective, such as finding out why a disk is full. Scout runs a series of commands through a small agent on the device, shows you each one, waits for your approval before anything risky, and writes a report to the ticket when it finishes.
Limit: Scout works on Windows devices only. In the device list, anything else is shown as "(not Windows)" and cannot be selected. There are no macOS or Linux agents.
Before you start
- Starting and approving Scout sessions is available to administrators, service managers and technicians.
- The device must be an active workstation, laptop, server or virtual machine in Inventory under the ticket's company, with a Windows operating system recorded.
- Scout needs the AI gateway. See Deployment. On a trial installation, Scout runs with stand-ins and the console shows MOCK MODEL and SIMULATED AGENT tags. The commands and output in that case are simulated.
- In production you need a way to run a command on the device as an administrator, for example a remote shell from your RMM tool.
Start a session
- Open the ticket and find the Scout diagnostics section.
- Click New Scout session.
- Choose the Device. Devices on the ticket and devices that belong to the ticket's contact are labelled.
- Write the Objective: what Scout should find out and what the report should contain. Be specific. "Why is D: full on this machine; list the largest folders and say whether the logs can be archived" works better than "check disk".
- Choose a Profile (see below).
- Set the Time budget in minutes. The default is 30; the range is 2 to 120.
- Click Start Scout.
You can also start a session by applying a Scout diagnostic suggestion from AI triage.
Profiles
| Profile | What Scout may do |
|---|---|
| Read-only | Read information only. Nothing is changed on the device. Tools that change anything are refused. |
| Standard | Reads, plus two safe fixes that run without asking: restarting a service from a fixed list, and clearing old temporary files. Scripts wait for your approval. |
In both profiles, reading a file or registry location that looks like it holds credentials waits for approval.
Connect the device
In production, Scout does not reach out to the device. After you click Start Scout, the ticket shows a box headed Run this on the device with a one-line PowerShell command.
- Click Copy.
- Open an administrator shell on the device, for example your RMM tool's remote shell.
- Paste the command and run it.
- Click Open console on the ticket.
The command contains a token for this session only, and it is shown once. If the agent has not checked in within the allowed time (15 minutes unless your administrator has changed it), the session closes and you need to start a new one.
Limit: Frontline PSA cannot push the agent to a device through your RMM. Someone has to run the launcher command on the device.
The console
The console opens at Intelligence → Scout and updates live. The header shows:
- The device, the session status and the profile.
- The objective, the company, who started the session, and when the agent was last seen.
- Three budgets with bars: Turns, Commands and Minutes. When any budget runs out, Scout is asked for its final report with what it has.
Approve, edit or deny a command
When Scout wants to run something that needs approval, a card headed Approval needed appears at the top of the console. It shows the tool, its class, the reason it was classed that way, Scout's explanation and the exact arguments.
| Class | Examples | Needs approval |
|---|---|---|
| Read | System information, processes, services, disk usage, event logs, files, registry, network probes | No |
| Safe write | Restart an allowed service, clear old temporary files | No (Standard profile only) |
| Sensitive read | A file path or registry hive that looks like credentials | Yes |
| Destructive | Running a script | Yes |
| Broad impact | A script that reboots, formats, uninstalls or deletes widely; any tool Scout does not recognise | Yes, and you must type CONFIRM |
For each card you can:
- Click Approve to run the command as shown.
- Click Edit…, change the arguments, then click Approve with edits. Edited arguments are classified again. If your edit makes the command more dangerous, the stricter rule applies.
- Click Deny. Scout is told and carries on without it.
Add a Reason if you want one kept with the decision. If nobody decides within 15 minutes, the command is not run.
Guide the session
While the session is running you can type in the box at the bottom, for example "the user says it started after Tuesday's update", and click Send. Scout takes it into account on its next step.
Click Stop to end the session at any time.
Tabs
- Commands lists every command with its class, exit code and duration. Click one to see the arguments as run and the full output. Approved and denied commands show who decided.
- Ledger is the complete record of the session: creation, the agent connecting, each model turn, each command queued, approved, denied, picked up and finished, your messages, and the close.
- Report is Scout's final report.
How a session ends
| Status | Meaning |
|---|---|
| Waiting for agent | The launcher has not been run yet, or the agent has not connected. |
| Running | Scout is working. |
| Objective met | Scout answered the objective. |
| Could not proceed | Scout could not answer the objective and says why. |
| Budget spent | A budget ran out. The report covers what was found so far. |
| Expired | The agent did not connect in time. |
| Stopped | Someone clicked Stop. |
| Failed | The session ended on an error. |
When a session ends, the report is written to the ticket as an internal note, word for word, with a table of the commands that were run as evidence. The console shows which note it wrote.
The ledger
Every event in a session is added to a ledger in which each entry is chained to the one before it by a hash. Changing or removing an entry afterwards breaks the chain. When a session has ended, the console checks the chain and shows ledger intact or ledger BROKEN. The short code at the end of each ledger line is that entry's link in the chain.
Find past sessions
- On a ticket, Scout diagnostics lists the sessions started from it, with status and the number of commands.
- Intelligence → Scout lists every session with its device, ticket, company, objective, profile and who started it. Sessions that are waiting on an approval are flagged.