Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
Arize AI avatar

Claude Code Tracing

  • 30 repo stars
  • Updated August 4, 2026
  • Arize-ai/coding-harness-tracing

Automatically trace Claude Code sessions to Arize AX or Phoenix using OpenInference spans for agent observability and debugging.

About

claude-code-tracing automatically instruments Claude Code sessions, emitting OpenInference spans to Arize AX or Phoenix. It gives teams observability into agent runs so they can analyze and debug LLM/agent behavior. Useful for monitoring and evaluating coding-agent sessions in production or development.

  • Automatic Claude Code tracing
  • OpenInference spans
  • Arize AX and Phoenix support
  • Agent observability

Claude Code Tracing by the numbers

  • Data as of Aug 5, 2026 (Skillselion catalog sync)
/plugin marketplace add Arize-ai/coding-harness-tracing
/plugin install claude-code-tracing@coding-harness-tracing

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
repo stars30
Last updatedAugust 4, 2026
RepositoryArize-ai/coding-harness-tracing

What it does

Automatically trace Claude Code sessions to Arize AX or Phoenix using OpenInference spans for agent observability and debugging.

README.md

Claude Code Tracing

Automatic OpenInference tracing for the Claude Code CLI and the Claude Agent SDK. Spans are exported to Arize AX or Phoenix.

Setup

The installer prompts for your backend (Phoenix or Arize AX) and project name, writes credentials to ~/.arize/harness/config.yaml, and registers the hooks in ~/.claude/settings.json.

Pass --with-skills to also symlink the manage-claude-code-tracing skill into the current directory's .agents/skills/ so Claude can help you manage the configuration interactively.

Claude Code marketplace

The marketplace flow registers the hooks but skips the interactive wizard, so backend credentials and content-logging preferences must be set directly in ~/.claude/settings.json under env:

{
  "env": {
    "ARIZE_PROJECT_NAME": "claude-code",
    "ARIZE_API_KEY": "<your-arize-api-key>",
    "ARIZE_SPACE_ID": "<your-arize-space-id>",
    "ARIZE_LOG_PROMPTS": "true",
    "ARIZE_LOG_TOOL_DETAILS": "true",
    "ARIZE_LOG_TOOL_CONTENT": "true"
  }
}

For Phoenix, swap the Arize keys for PHOENIX_ENDPOINT (and optional PHOENIX_API_KEY). Each ARIZE_LOG_* flag accepts "true" or "false" — set to "false" to opt out per category. Env values take precedence over ~/.arize/harness/config.yaml.

# Install
claude plugin marketplace add Arize-ai/coding-harness-tracing
claude plugin install claude-code-tracing@coding-harness-tracing

# Uninstall
claude plugin uninstall claude-code-tracing@coding-harness-tracing
claude plugin marketplace remove Arize-ai/coding-harness-tracing

Remote setup

macOS / Linux:

# Install
curl -sSL https://raw.githubusercontent.com/Arize-ai/coding-harness-tracing/main/install.sh | bash -s -- claude

# Uninstall
curl -sSL https://raw.githubusercontent.com/Arize-ai/coding-harness-tracing/main/install.sh | bash -s -- uninstall claude

Windows (PowerShell):

# Install
iwr -useb https://raw.githubusercontent.com/Arize-ai/coding-harness-tracing/main/install.bat -OutFile $env:TEMP\install.bat
& $env:TEMP\install.bat claude

# Uninstall
iwr -useb https://raw.githubusercontent.com/Arize-ai/coding-harness-tracing/main/install.bat -OutFile $env:TEMP\install.bat
& $env:TEMP\install.bat uninstall claude

Local setup

git clone https://github.com/Arize-ai/coding-harness-tracing.git
cd coding-harness-tracing

macOS / Linux:

# Install
./install.sh claude

# Uninstall
./install.sh uninstall claude

Windows:

# Install
install.bat claude

# Uninstall
install.bat uninstall claude

Default Settings

Setting Default
Harness key claude-code
Project name claude-code
Phoenix endpoint http://localhost:6006
Arize AX endpoint otlp.arize.com:443
Hook config file ~/.claude/settings.json
Hook events registered SessionStart, SessionEnd, UserPromptSubmit, UserPromptExpansion, PreToolUse, PostToolUse, PostToolUseFailure, Stop, StopFailure, SubagentStart, SubagentStop, Notification, PermissionRequest, PermissionDenied, PreCompact, PostCompact
State directory ~/.arize/harness/state/claude-code/
Log file ~/.arize/harness/logs/claude-code.log

Verifying tracing

Run any Claude Code session as you normally would (e.g. claude or claude -p "hello"). The installed hooks fire on every SessionStart, UserPromptSubmit, PreToolUse, etc.

  • Errors and ARIZE_VERBOSE=true activity land in ~/.arize/harness/logs/claude-code.log. To see routine hook activity (session_start fired, emitted LLM span, etc.), add "ARIZE_VERBOSE": "true" under env in ~/.claude/settings.json and re-run a session.
  • Confirm spans appear in your configured project in Arize AX or Phoenix.
  • Set "ARIZE_TRACE_ENABLED": "false" under env to temporarily disable tracing without uninstalling, or "ARIZE_DRY_RUN": "true" to build spans without sending them.

See the main README's Environment variables section for the full list of runtime overrides.

Related skills

AI & Agent Buildingagentsautomation

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.