Skip to main content
Version: 1.0.0

Quick Start

Use device-monitor-agent to walk the whole loop once: install, sign in, import the Agent, create its Task, run it, and read the reports.

Before you begin​

First make sure the environment is fully installed.

Download the two files of device-monitor-agent from the IndustrialClaw repository:

FileWhat it is
device-monitor-agent.zipThe Agent. Its device-monitor Skill takes read-only snapshots of CPU, temperature, and disk.
monitor-local-cpu-task.mdThe Task. It samples CPU usage, CPU temperature, and the root disk, and writes one report for each.

1. Install and launch​

Run the installer as described in Installation, then double-click the IndustrialClaw desktop shortcut.

The launcher checks the environment, starts the containers, waits up to 60 seconds for the Dashboard to answer, and opens the browser for you.

View: IndustrialClaw launcher starting the platform

IndustrialClaw launcher starting the platform

The browser shows a warning on the first connection. Open the advanced options and continue.

View the certificate warning

Browser certificate warning

2. Sign in and set a password​

Open https://<device-ip>:18765/ and sign in with the default credentials admin / admin.

View the sign-in screen

IndustrialClaw sign-in page

The Dashboard immediately requires a password change. The new password must be 6–128 characters and contain uppercase letters, lowercase letters, and numbers. Sign in again with the new password.

View: First sign-in password change

First sign-in password change

note

The admin account created at install time is the bootstrap admin. It is the only account that can delete other accounts or force another account to sign out. See Roles and limits.

3. Import the Agent​

  1. Open Dashboard → Agent Management and select Create Agent.
  2. Under Import Package, choose device-monitor-agent.zip and select Next.
  3. Under Security Scanner, review the device-monitor Skill's risk, file locations, permissions, mounts, and ports.
  4. Acknowledge any finding that requires it, then complete Create Agent.
View: Import Package step

Import Package step

View the security scan

Security Scanner results

View the created Agent

Agent created

device-monitor-agent now appears in the list with no Tasks yet—that is expected. Tasks are created separately, in the next step.

4. Create the Task​

  1. Under Task Management, select Create Task.
  2. Choose device-monitor-agent as the Agent, then monitor-local-cpu-task.md as the Task Markdown.
  3. Leave Material ZIP empty—this Task reads the device directly.
  4. Select Create, then open the new row under Available Tasks to confirm the definition stored on the device.
View: Create Task dialog

Create Task dialog

The Task runs two rounds (cycle: 2) and retries a failed round once (retry: 1). Each round runs three steps, each sampling five times:

StepWhat it samplesReport
1CPU usagecpu-usage-monitor-report.md
2CPU temperaturecpu-temperature-report.md
3Root disk usage and free spacedisk-monitor-report.md

The second round appends to the same reports. Final Verification then runs verify_report.py on all three reports.

5. Start the Task​

Select START on the monitor-local-cpu-task row. The button briefly shows STARTING…, and a new record appears under Task Runs.

View: Starting a Task

Starting a Task

START creates, claims, and dispatches the run in one call, so it never waits in a queue: the record appears as running and ends as done or failed.

View a running Task

Task running

View a completed Task

Task done

You do not start an Agent container by hand. START creates, claims, and dispatches the run, and writes an Audit record naming who started it.

6. Read the result​

Select the empty area of the Task Run row—not the STOP button—to open the detail dialog. It has three tabs:

TabWhat it holds
LogThe latest 400 lifecycle, round, step, and error records
SessionThe USER, ASSIST, TOOL, and RESULT messages exchanged with the model
OutputsThe three reports, available for download
View: Task Run outputs

Task Run outputs

A value the device cannot read, such as a CPU temperature with no sensor, is written as Unknown rather than estimated.

The run reaches done only when Final Verification passes and all three reports exist as regular files. The platform verifies those files; it never creates, copies, or moves them.

Next steps​