--- title: Agents description: The per-tenant remote-support view at --- # Agents ## Purpose AGENTS shows one tenant's support agents: the preflight readiness checklist, last-known telemetry, and the enrolled-agent table with the gated session actions. It is scoped to the tenant selected on [Tenants](/docs/console/tenants) (a tenant picker plus LOAD button at the top can change it). Deep link: `ops.useargus.co/#/agents` (nav group REMOTE SUPPORT). ## What it shows * Readiness (right rail): a READY or NOT READY verdict, the individual preflight checks with hints on failures (feature flag, operator CIDRs, ops token, tenant consent, agent online), and fleet counts (enrolled, active, online, revoked). All checks must be green before Tunnel In will succeed. * ACTIVITY (last 24h): activations, activation fails, commands, command fails, enrollments, auth fails, revocations, and last activity. * AGENT TELEMETRY: last-known state pushed by each agent over its support channel, shown even between sessions (last-known, not live). Columns: AGENT, VERSION, UPTIME, CAMERAS, CPU, MEM, PIPELINE, LAST SEEN, plus optional sub-rows for the last error, a log tail (up to 20 lines), and network details (local IP, subnet, gateway, DNS). * TUNNEL-IN / ACTIONS: copy-ready argusctl commands for the gated actions (enroll ticket, activate, view\_camera, revoke), which run from a tailnet machine. * AGENTS: the enrolled-agent table: AGENT, NAME, STATUS, CONNECTION (online, stale, offline), LAST SEEN, INSTANCE, and per-agent action buttons. ## Actions Active agents carry three buttons: * Tunnel In opens a guided modal. When the one-click proxy path is available it takes a reason and offers a one-click "Activate tunnel" button; otherwise (or as the fallback) it shows the exact `argusctl support activate` command and polls the session status every 3 seconds until the tunnel is established. If one-click is not enabled server-side, the modal says so and falls back to the CLI path. * View Camera opens a modal that templates the `argusctl support command --cmd view_camera` command for a camera id; you run it from a tailnet machine and paste the result (its `image_b64`) back into the modal to render the frame in the console. The browser cannot pull a frame directly by design. * REVOKE is destructive: a step-up modal requires typing the agent id and a reason, warns the agent must re-enroll to reconnect, and cannot be undone. It is inert until the server-side browser-commands flag is enabled (the modal reports "Browser commands not enabled yet"). ## Matching CLI ```bash argusctl agent ls --tenant argusctl agent telemetry --tenant argusctl agent activate --tenant --reason "investigate alert" argusctl agent view-camera --tenant --session --camera --out frame.jpg argusctl agent revoke --tenant --reason "stale enrollment" --yes ``` See [argusctl agent](/docs/cli/agent), [argusctl support](/docs/cli/support), and the [remote-in guide](/docs/remote-support/remote-in). ## Notes * Polled on the 15 second timer. The four data sources (readiness, metrics, telemetry, agents) are fetched independently, so one failing never blanks the others. * The tenant toolbar lives outside the refreshed panel, so typing is never interrupted by a tick. * Until an agent has enrolled a passkey, activation returns "agent offline" by design. --- For a semantic overview of all documentation, see [/sitemap.md](/sitemap.md) For an index of all available documentation, see [/llms.txt](/llms.txt) For agent-facing discovery, including API and MCP surfaces, see [/agents.md](/agents.md)