As AI coding agents spread across terminals, SDKs, and IDEs, platform teams need answers to practical questions: which agents are adopted, what they cost in tokens, where tool calls fail, and how activity ties to pull requests. Dynatrace expands AI Coding Agent Monitoring for Claude Code, Gemini CLI, Codex CLI, and related tools, ingesting OpenTelemetry signals for sessions, cost, tools, errors, and latency in one observability layer.

📑Table of Contents
  1. What Dynatrace Adds for Coding Agent Observability
  2. Supported Agents and OpenTelemetry Integration
  3. Why AI Agent Observability Matters Now
  4. OpenTelemetry Setup and Public Examples
  5. A CLI-First Workflow Perspective
  6. Frequently Asked Questions
  7. Summary

What Dynatrace Adds for Coding Agent Observability

According to Dynatrace’s announcement, supported experiences include Claude Code, Google Gemini CLI, OpenAI Codex CLI, OpenCode, and GitHub Copilot SDK. Instead of stitching fragmented telemetry across workflows, teams can track adoption, cost signals, reliability alerts, and engineering outcomes such as commits and PRs in one place.

Agents that emit OpenTelemetry directly—Claude Code, Gemini CLI, Codex CLI, and OpenCode—send session, token, cost, tool, error, and latency data to Dynatrace. For embedded GitHub Copilot SDK workflows, Dynatrace adds production context, software delivery automation, and GitHub integrations so agent paths can be traced closer to real delivery pipelines.


Supported Agents and OpenTelemetry Integration

AgentOTelWhat Dynatrace surfaces
Claude CodeBuilt-in, no code changesSessions, tokens, cost, tools, API health, commits/PRs
Gemini CLIOTel + prebuilt dashboardsAgent activity correlated with platform signals
Codex CLIOpt-in OTelCross-surface audit, traces, token alerts
OpenCodeStandard OTLP env varsLLM volume, sessions, tools, latency
GitHub Copilot SDKEmbedded (Dynatrace adds context)Execution paths, logs, policy-relevant events

Source: Dynatrace official blog (June 2026)


Why AI Agent Observability Matters Now

Organizations are past the “are coding agents useful?” phase. The focus shifted to safe scale-up, governance, and measurable impact. When one team lives in Claude Code, another embeds Copilot SDK, and others pilot Gemini CLI or Codex CLI, cost spikes and silent tool failures hide in local logs. Dynatrace positions a single layer for adoption visibility, spend signals, and reliability regressions.

The official post quotes a customer who could not break down model usage and cost before instrumenting Claude Code, then identified inefficient model choices afterward. That narrative matches teams moving from individual experimentation to managed rollout.


OpenTelemetry Setup and Public Examples

Dynatrace publishes dynatrace-ai-agent-instrumentation-examples on GitHub with step-by-step OpenTelemetry export, Dynatrace ingestion, and dashboard analysis for supported coding-agent workflows. Claude Code’s built-in OpenTelemetry lowers the setup barrier.

Codex CLI monitoring is opt-in, suited to audit and governance across CLI, IDE, and app surfaces. OpenCode routes telemetry via standard OTLP environment variables, keeping terminal-first agents on the same strategy.


A CLI-First Workflow Perspective

The author previously relied on MCP but now prefers CLI tools where agents consume less context and handle commands more reliably—while still using MCP where no CLI exists. Splitting design in Claude Code and implementation/review in Codex is common; long-running tasks make it hard to see which subtasks burn tokens.

Dynatrace-style AI Agent Observability helps explain team-level cost and failure patterns even when individuals juggle ChatGPT Pro ($100) and Claude Max ($200) plans. Combined with the Dynatrace MCP Server for production context inside the IDE, it connects agent decisions to delivery and production signals—not just log aggregation.


Frequently Asked Questions

Q: Does Claude Code monitoring require code changes?

A: Dynatrace states Claude Code includes built-in OpenTelemetry, so metrics and logs can reach Dynatrace without code changes.

Q: Is Codex CLI monitored by default?

A: Codex CLI uses opt-in OpenTelemetry. After enabling it, teams can alert on traces, failures, and rising token use in Dynatrace.

Q: Why is GitHub Copilot SDK treated differently?

A: Embedded SDK agents need production context and GitHub-linked traces rather than terminal-only OTel export.

Q: Where are configuration examples?

A: See dynatrace-ai-agent-instrumentation-examples on GitHub for OTel export and dashboard walkthroughs.

Q: Is this still useful in MCP-only setups?

A: Session visibility helps anywhere, but Dynatrace emphasizes cross-model correlation with PRs and production—most valuable when CLI and SDK agents coexist.


Related articles: GitHubがClaude・Codexなど第三者コーディングエージェントのセキュリティ検証をGA化Codex app 26.609:リセット貯金・Developer mode・Browser Use高速化が追加SupabaseがAIコーディングエージェント向けプラグインを発表:MCPとAgent Skillsを同梱

Summary

Dynatrace extends AI Coding Agent Monitoring across Claude Code, Gemini CLI, Codex CLI, and more, unifying OpenTelemetry signals for sessions, cost, tools, and errors. As agent adoption becomes an organizational concern, that unified view supports governance and measurable engineering outcomes—worth considering for CLI-first workflows that need accountability on long agent runs.

Related new article:

krona23

Author

krona23

Over 20 years in the IT industry, serving as Division Head and CTO at multiple companies running large-scale web services in Japan. Experienced across Windows, iOS, Android, and web development. Currently focused on AI-native transformation. At DevGENT, sharing practical guides on AI code editors, automation tools, and LLMs in three languages.

DevGENT about →

Leave a Reply

Trending

Discover more from DevGENT

Subscribe now to keep reading and get access to the full archive.

Continue reading