Kiro

Trace Kiro CLI agent turns, tool calls, and credit usage in Phoenix.

Trace Kiro CLI agent turns, tool calls, and credit usage in Phoenix for full observability.

Trace your Kiro CLI sessions in Phoenix with the coding-harness-tracing toolkit — every turn shows up as a trace with the model call, each tool it runs, credit cost, and turn duration, grouped into its session. No application code changes required: the toolkit registers hooks on a Kiro agent and streams OpenInference spans to Phoenix.

Launch Phoenix

The fastest way to get started with Phoenix is by signing up for a free Phoenix Cloud account. If you prefer, you can also run Phoenix in a notebook, self-host it, or use it directly from your terminal.

Go to the settings page in your Phoenix instance to find your endpoint and API key. A self-hosted Phoenix defaults to http://localhost:6006; the API key is only required when auth is enabled.

Install

The curl installer is the simplest — it runs a short wizard that saves your Phoenix credentials for you. Use a local clone if you’d rather run the installer from a checkout of the source.

Curl installer (recommended)

macOS / Linux:

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

Windows (PowerShell):

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

Local clone

git clone https://github.com/Arize-ai/coding-harness-tracing.git
cd coding-harness-tracing
./install.sh kiro          # macOS / Linux
install.bat kiro           # Windows

The installer prompts you through:

1. Agent name — the Kiro agent to install hooks into (default: arize-traced)
2. Set as default — whether to run kiro-cli agent set-default <name> so the agent is used by default
3. Backend — select Phoenix and enter your endpoint and optional API key
4. Project name — project name in your backend (default: kiro)
5. User ID — optional user identifier added to all spans
6. Logging — whether to include prompt text, tool content, and tool details in spans

Once installed, run Kiro as usual. If you set the traced agent as the default during install:

kiro-cli chat

Otherwise, point Kiro at the traced agent explicitly:

kiro-cli chat --agent arize-traced

Configuration

Credentials live under harnesses.kiro in ~/.arize/harness/config.json (written by the installer). Environment variables override values in config.json and can be set in your shell profile so they apply to every Kiro session.

export PHOENIX_ENDPOINT="http://localhost:6006"
export PHOENIX_API_KEY="<your-api-key>"   # optional, only if auth is enabled
export PHOENIX_PROJECT="kiro"
export ARIZE_TRACE_ENABLED="true"

On the Phoenix backend, set the project name with PHOENIX_PROJECT (or PHOENIX_PROJECT_NAME); ARIZE_PROJECT_NAME is Arize-only and ignored here. ARIZE_TRACE_ENABLED is a backend-agnostic harness setting and keeps the ARIZE_ prefix regardless of destination.

Redaction controls

Each ARIZE_LOG_* flag accepts "true" or "false" and defaults to "true". Set to "false" to opt out per category:

export ARIZE_LOG_PROMPTS="false"
export ARIZE_LOG_TOOL_DETAILS="false"
export ARIZE_LOG_TOOL_CONTENT="false"

Flag	Redacts
ARIZE_LOG_PROMPTS	User prompt and assistant response text
ARIZE_LOG_TOOL_DETAILS	Tool names and arguments
ARIZE_LOG_TOOL_CONTENT	Tool call output content

The same flags can be set in config.json under harnesses.kiro.logging as log_prompts, log_tool_details, and log_tool_content — env vars take precedence.

Observe

Once tracing is enabled, Kiro activity is streamed to Phoenix. You’ll see:

• Turn traces — each agent turn (user prompt to assistant response) as a parent LLM span
• Tool call spans — one per pre/post tool event pair, parented to the LLM turn
• Session grouping — all turns from the same Kiro session grouped by session.id
• Credit cost — Kiro meters in credits rather than tokens; cost is captured as kiro.cost.credits
• Model and duration — llm.model_name, kiro.turn_duration_ms, and kiro.context_usage_percentage from the session sidecar

Span attributes

LLM spans are enriched from the session sidecar at ~/.kiro/sessions/cli/<session_id>.json with model name, cost in credits, metering usage, and turn duration. Enrichment is fail-soft — if the sidecar is unavailable, the span is emitted with basic attributes only.

Attribute	Description
session.id	Kiro session UUID
llm.model_name	Model ID from the session sidecar
kiro.cost.credits	Cost in credits from metering data
kiro.metering_usage	Full metering usage JSON
kiro.turn_duration_ms	Turn duration in milliseconds
kiro.agent_name	Name of the Kiro agent
kiro.context_usage_percentage	Context window usage percentage

Known limitations

• Token counts are 0. Kiro CLI does not report prompt or completion token counts in current versions and meters in credits instead. Token count attributes are omitted when 0; see kiro.cost.credits.
• FIFO tool matching. Kiro does not expose a tool-call ID, so pre/post tool events are matched using a FIFO stack. This assumes serial tool execution within a session.
• Per-workspace agents not supported. Only global agents under ~/.kiro/agents/ are instrumented.

Reference

For the full list of environment variables, default file paths, and troubleshooting steps, see the Kiro tracing README.

Uninstall

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

Uninstall removes hook entries from every Kiro agent file. If the arize-traced agent was created by the installer, that agent file is deleted; pre-existing agents are preserved.

Resources

Arize Coding Harness Tracing

OpenInference

Kiro Documentation