Tools

Once connected, your assistant can call these tools:

Alerts

Tool	What it does
list_alerts	Browse and filter alerts — by severity, type, state, hostname, text, or time window — with pagination.
get_alert	Full detail for one alert, optionally including its process-level event timeline.
get_alert_summary	Alert counts broken down by severity, type, and state.
get_alert_trend	Per-severity alert counts in time buckets — how alert volume is trending over a window.

GPU fleet

Tool	What it does
list_gpus	The GPU fleet inventory — one row per GPU with identity, health, and freshest telemetry.
get_gpu	Full detail for one GPU: specs, security summary, XID errors, live telemetry, processes.
get_fleet_summary	The fleet dashboard payload: alert histograms, top alert types, asset counts, aggregate telemetry.

Utility

Tool	What it does
list_scopes	The customers, workspaces, and API keys you can scope queries to.
get_docs	This documentation, so the assistant can ground its answers.

All tools are read-only — an MCP client can't change anything in your account.

Filter values (severities, alert types, states) match the REST API's — see Filtering. Tool failures return the same { "error", "message" } envelope as the REST API.