Skip to content

AI Agents — TrackMe Agentic

A fleet of AI advisors that inspect your monitored estate, propose grounded configuration changes, and — with your explicit consent — apply them through TrackMe's own REST API.

Key Capabilities

Six Specialist Advisors

ML Advisor (outlier models), Feed Lifecycle (DSM/DHM thresholds), FLX Threshold, FQM (field quality & data dictionary), Component Health (WLK/MHM), plus the catalog-driven Concierge — each scoped to one monitoring surface.

Inspect or Act

Every advisor runs read-only by default and surfaces typed, reviewable recommendations. In act mode it applies the fix and verifies every write. Write tools are simply absent until you opt into act.

One-Click from the Chat

Describe a symptom in the AI Assistant and it proposes the right advisor as an inline consent card — launched in place, with a live tool-call feed and a structured result. The chat never writes directly.

The AI Concierge

A catalog-driven generalist grounded in the live TrackMe REST API. Zero static write tools — every action is a consent-card click, with typed confirmation for destructive operations and RBAC enforced at the boundary.

Unattended Automation & Audit

Run the advisors on a per-tenant nightly schedule, scoped by your own priority and freshness filters. Every run — interactive or scheduled — is recorded in dedicated audit dashboards.

Trust, Auditability, Traceability

Decisions are typed and grounded in real data, never free-text guesses. Tag-filtered write gates, a decommission guard off by default, and an [AI Agent] audit trail on every change keep the agents reviewable.

AI Agents — TrackMe Agentic — 1
AI Agents — TrackMe Agentic — 2
AI Agents — TrackMe Agentic — 3

Deep Dive

Explore the key aspects of AI Agents in detail.

The big picture

From assistant to agents

TrackMe has always detected data availability and quality problems. With TrackMe Agentic, it now helps you understand and resolve them. The AI Assistant was the conversational front door; the AI Agents are the layer that acts — reading the same KV Store and search indexes a human analyst would, proposing the same moves, and applying them through the very same REST API a UI click would invoke.

What the advisors do

Every advisor follows one loop: observe the entity’s live context, metric history, scoring and audit trail; diagnose the evidence against a typed set of scenarios; recommend structured, reviewable changes; and — in act mode — remediate through TrackMe’s REST API, verifying each write. Five specialists cover the operational surface (ML models, DSM/DHM thresholds, FLX thresholds, field-quality dictionaries, and WLK/MHM health); the AI Concierge handles everything else by grounding itself in the live API catalog.

Safe by design

The agents are built to be reviewable, not opaque. They run inspect (read-only) by default; write tools are absent until you explicitly opt into act mode. Every write is RBAC-checked at the REST boundary and tagged with an [AI Agent] reason in the entity audit trail, and two dashboards record every run. You stay in control: nothing happens without your consent, and you can always answer what changed, who triggered it, and why.

Bring your own LLM — fully opt-in

TrackMe is provider-agnostic: connect OpenAI, Azure OpenAI, Anthropic, Google Gemini, Mistral, xAI, self-hosted Ollama, or Splunk Hosted (SLIM API). Nothing is sent anywhere until you configure a provider. TrackMe remains fully compatible with all supported Splunk releases — the AI Assistant and AI Concierge run on Python 3.9+, while the AI Advisors use the Splunk Agent SDK and require Splunk 10.2.x and later (Python 3.13). TrackMe enforces this automatically, so the agentic features are safe to roll out everywhere.

Ready to get started?

Request a free 90-day trial with all features enabled. No restrictions.