Observe Platform (Unified Observability Workspace)
A themeable (dark/light) observability workspace embedded natively in the admin console as an expandable Observe Platform sidebar menu. Rather than exporting telemetry to a separate, separately-licensed observability stack, AgentWatch surfaces its own operational picture in one place, org-scoped and access-controlled by the same session that governs the gateway.
| Surface | What it shows |
|---|---|
| AI Observability | Model / token / cost / latency KPIs, prompt lineage (with a no-raw-PII guarantee), agent topology, and quality evals |
| Application Observability | Distributed-trace waterfall and service-flow map |
| Infrastructure & Fleet | Honeycomb health grid — one hex per deployed agent, colored by health |
| Optimize | AgentKeel direct-vs-optimized comparison and per-agent savings |
| Business | Spend funnels by provider, team, and model |
| Security & Threat | Risk-ranked DLP / WAF events |
| Insider Risk | AI-agent actor risk scores, indicators, incidents, and enforcement |
| Software Delivery | Release guardian pass / warn / fail gates |
| Log Analytics | Query bar and volume histogram over recent activity |
Reachable at /{org-slug}/admin/dashboard → Observe Platform. Backed by org-scoped endpoints under /admin/api/observability. See §7.19.