Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tracectrl.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

The TraceCtrl Dashboard is a React + Vite + TypeScript SPA. Access it at http://localhost:3000 after starting the stack (port 3001 in dev mode). The sidebar is grouped into Monitor, Security, and Configure sections. / redirects to /agents.

​
Routes

PathSectionPage
/—Redirect to /agents
/agentsMonitorAgent inventory
/sessionsMonitorSessions list
/sessions/:traceIdMonitorTrace detail
/topologyMonitorTopology graph
/alertsSecurityGuardrail violations feed
/guardrailsSecurityRegistered guardrails
/scanSecurityStatic scan report
/riskSecurityRisk dashboard
/attacksSecurityAttack paths
/settingsConfigureSettings (incl. TraceCtrl Guards config)
/setup(not in sidebar)Onboarding

​
Monitor

​
Agents

Agent inventory view showing all discovered agents and their metadata.
  • Agent Table — name, framework, maturity level, observation count, and tool list.
  • Tool Expansion — click any agent row to expand and see all observed tools with their risk categories.

​
Sessions

Trace explorer. The list page (/sessions) shows one row per trace; clicking a row navigates to /sessions/:traceId for the detail view.
  • Trace Tree — vertical tree with icons per span type, indented connectors, expand/collapse, and duration labels.
  • Span Detail Panel — click any node in the tree to see timing, input/output values, grouped attributes (TraceCtrl, OpenInference, etc.), and resource metadata.
  • Waterfall Column — proportional duration bars in the session table for at-a-glance comparison.
  • Sortable Table — sort by root span name, span count, duration, or time.

​
Topology

Interactive graph visualization built with Cytoscape.js and dagre layout.
  • Agent nodes (blue) — each discovered agent with name, framework, and maturity.
  • Tool nodes (green) — each tool connected to its owning agent.
  • Edges — solid lines for agent-to-agent delegation, dashed lines for agent-to-tool usage.
  • Attacker View Toggle — switch to an attacker perspective that overlays exploitable chains in red.
  • Phase Overlay — “Show Phases” groups concurrent agents into execution phases with sequential “then” arrows.
  • Replay Scrubber — drag the timeline slider to highlight which agents are active at any point in time.
  • Project Selector — sidebar dropdown filters all pages by service name.
  • Light/Dark Theme — toggle in the sidebar footer, persists across sessions.
  • Click any node to see its full metadata in the sidebar panel.

​
Security

​
Alerts

Live feed of guardrail violations (failing tracectrl.guardrail.evaluation spans). The sidebar shows an unread-count dot when there are new violations since last visit. Server-Sent Events push new rows as the engine ingests them.

​
Guardrails

Registry of all guardrails that have been registered by the SDK (both in-process LLM judges and Protector Plus sub-guardrails). Each row shows agent, name, severity, mode (monitoring/blocking), timing (pre_input/post_output), provider (judge_llm/protector_plus), health, and 24h activity.

​
Scan Report

Output of the static OpenClaw analysis run from /api/v1/scan. Shows per-check pass/fail with findings and remediation, plus a static topology snapshot.

​
Risk Dashboard

CISO-level risk overview powered by the TAGAAI attack graph engine.
  • System Risk Score — single 0-10 score with severity badge (Critical / High / Medium / Low).
  • Stat Cards — critical paths count, agents at risk, learning agents (< 10 observations).
  • Per-Agent Risk Table — sortable table showing each agent’s highest risk score, severity, attack path count, and top triggering rule.
  • Computed At — timestamp of the last risk computation cycle.

​
Attack Paths

Ranked list of exploitation chains detected by TAGAAI rules.
  • Path List — ordered by risk score descending; each entry shows rule name, OWASP category, severity badge, and involved agents.
  • Expandable Step Chains — click any path to reveal the full step-by-step attack chain from entry point to impact.
  • Service Filter — filter paths by service name.

​
Configure

​
Settings

Application settings, including the TraceCtrl Guards / Protector Plus configuration UI: endpoint URL, API key, and which of the seven Protector Plus sub-guardrails (llm, keyword, regex, pii, vector, content_moderation, system_prompt_protection) are enabled. Values persist to the protector_plus_config table and are fetched by the SDK at startup.

​
Development

cd tracectrl/ui
npm install
npm run dev   # Vite dev server at :5173 (mapped to host :3001 via docker compose dev)
The dashboard reads the engine URL from VITE_ENGINE_URL (default: http://localhost:8000). This is baked at build time — for Docker, it’s set as a build ARG.