bmad-observability-agent

B-MAD Observability Agent

A comprehensive OpenTelemetry observability expert agent for the B-MAD Method (Breakthrough Method for Agile AI Driven Development).

Overview

The B-MAD Observability Agent (O11y Engineer) is an AI-powered specialist that helps you build production-grade observability using OpenTelemetry. It acts as a Senior Observability Engineer embedded in your AI-assisted development workflow, guiding you from zero instrumentation to a fully validated, SLO-driven observability stack.

The agent integrates directly with the B-MAD multi-agent ecosystem, handing off epics and stories to Scrum Master (Bob), Developer (Amelia), and Test Architect (Murat) agents for implementation and validation.

Key Features

Observability Assessment – Comprehensive maturity scoring (0-100) with gap analysis and improvement roadmaps
Observability Spec Generation – Use-case-driven specifications that define trace, metric, and log contracts
Collector Configuration – Design and optimize OpenTelemetry Collector pipelines with best-practice processor ordering
Semantic Convention Validation – Create, validate, and manage conventions with OpenTelemetry Weaver, integrated into CI/CD
SLO Definition – Performance, reliability, and availability KPIs with approval workflows and test story generation
MCP Rules & IDE Integration – Project-aware Dynatrace DQL queries embedded in IDE rule files
Dynatrace Automation – Dashboards, notebooks, workflows, and alerting as code via dtctl
MCP-Powered Discovery – AI-driven environment discovery and context-aware DQL generation via the Dynatrace MCP server
Cross-Agent Handoff – Structured handoffs to sprint planning, development, and testing agents

The 8-Phase Workflow

The recommended path from zero to production-grade observability follows eight phases:

Phase	Name	Trigger	What It Does
1	Assessment	`assess-observability`	Scan current state, maturity score, gap analysis
2	Observability Spec	`generate-observability-spec`	Use-case-driven specs with companion epic
3	Collector Configuration	`configure-pipeline`	Resource detection, processors, PII cleanup
4	CI/CD with Weaver	`validate-semconv`	Automated semconv validation in CI/CD
5	SLO Definition	`define-slos`	Performance/Reliability/Availability KPIs
6	MCP Rules	`configure-mcp-rules`	IDE-specific rule files with DQL
7	Implementation	Handoff to sprint-planning	Bob plans, Amelia implements, Murat tests
8	Validation	`validate-traces`	Query validation gate, dual-format reports

See the Recommended Workflow page for the full overview and Mermaid diagram.

Getting Started

Installation Guide – Prerequisites, installation methods, verification
Quick Start Tutorial – From zero to observability in minutes

Workflow Phases

8-Phase Workflow Overview

Reference

Guiding Principles

The O11y Engineer operates according to these principles:

Three pillars in harmony – Traces, metrics, and logs must correlate
Instrument once, export everywhere – Vendor neutrality is key
Cardinality is the enemy – Always consider dimensional explosion
Semantic conventions are contracts – Respect them
Observability as code – Version control everything
Validate before you deploy – Test in staging first

Resources

isiobservable YouTube Channel – OpenTelemetry tutorials and deep dives
Dynatrace dtctl Documentation
Dynatrace MCP Server
OpenTelemetry Weaver

Need help? Open an issue or reach out on Discord.

This site is open source. Improve this page.