Dashboard capabilities

Day to day you use Estate, Traffic, and Work. Guardian watches that same activity in the background, opens findings in Work for high risks, and recommends the next action. Labs is on-demand red team and assessments, not the always-on review. Control plane covers Estate Discovery, Work Approvals / Flows, Policies Gitops, and Costs Imported.

Run curated adversarial prompts across categories (prompt injection, jailbreak, data exfiltration, system prompt extraction, PII). The UI includes scenario presets, optional asset scope, run history in the browser, comparison versus the previous run, JSON export of results, and links to custom rules and assessments for remediation.

Queue allowlisted probe profiles against your live gateway, once or on a recurring schedule (due runs are queued hourly). Quick Check proves Intertrace is on; heavy profiles (OWASP / Full) run NVIDIA Garak against the agent through Intertrace on a separate worker. Each run keeps a summary, the Garak artifacts, and an HTML report. Test keys for a run always expire.

Ingested events can include agent context (session, run, hop), tooling summaries and optional detail (when allowed), and per-rail timing. Each conversation is one runtime session; the session page opens with a trace viewer that shows every turn as a step tree with its decision, input/output, and effects. The activity ledger (agent page, session page, Runtime) lists what agents actually touched: each tool call as a verb (read, write, delete, send, deploy, and so on) on a resource such as a file, repository, table, or recipient.

Organizations and individual agents define tool execution policy (monitor vs block, allow/deny lists, approval rules). A policy pack adds a versioned template; organization and pack denies are additive, and an agent can tighten but never loosen what the organization sets. Configure it in Policies (Overview, Access, Data & attacks, Approvals, Detections, Policy as code) or on each agent's page.

Structured findings link classifier output, runtime checks, and gateway evidence. Incidents group work for triage. For automation, the product exposes exports such as hop-level JSON for an agent run (see dashboard copy on the Assessments page for the GET /api/exports/agent-run/… pattern).