Skip to main content

​
What is support agent?

The Support agent is the resolution engine of Endpoint Care: it takes an employee’s problem in plain language and fixes it autonomously on the device, with a guardrail on every action.

​
How it works

  1. The employee describes the problem in the app, or clicks a quick action (“Fix Email”, “Clean Disk”).
  2. The agent matches the request to a skill, the playbook for that problem type.
  3. It builds an ordered plan of tool calls and passes it through the guardrails: intent screening, allowed-tools validation, central policy.
  4. Steps execute one at a time on the device. Diagnostics are read-only; risky steps pause for a preview and the user’s consent.
  5. The agent verifies the outcome, asks the user to confirm, and records the whole run as a ticket with the full trail.
Every run is visible in the console with its plan, per-step results, gate decisions, and cost. See Runs.

​
What makes it different

  • It sees the real machine. Diagnostics read live system state: processes, disk, network, mail configuration. Not a guess from a ticket.
  • It reasons, within playbooks. The LLM adapts the fix to the specific situation, but only inside expert-written skills and only through deterministic, code-reviewed tools.
  • The user stays in control. Nothing risky happens without a preview and consent, and declining never breaks the run.

​
When it can’t fix something

The Support agent never dead-ends. If the fix is out of reach (server-side cause, missing permission, declined consent), it packages what it found and hands off: the run escalates with the diagnostic trail attached, and the Triage agent flow ensures it reaches the right team.