Skip to main content

MCP Agents

Use the Machine Control Protocol interface to let AI assistants and automation frameworks interrogate your telemetry.

Agent Lifecycle​

  1. Define agent personas and allowed tool invocations.
  2. Provide scoped credentials so each agent can only query approved tenants.
  3. Configure prompt templates that translate natural-language questions into OTLP queries.
  4. Capture feedback and reward signals to refine the agents over time.
  • Triage bot – Watches new incidents and proposes root-cause hypotheses.
  • Capacity planner – Periodically inspects historical metrics to recommend scaling adjustments.
  • Release guardian – Cross-checks new deploys against performance budgets and toggles feature flags when anomalies appear.

Review the SDK samples for code that wires MCP agents into Slack, Teams, and PagerDuty workflows.