AI CLI Tools — Overview

AI CLI Tools — Overview screenshot 1
AI CLI Tools — Overview screenshot 2
AI CLI Tools — Overview screenshot 3

Overview

A unified dashboard for Claude Code, Codex CLI, and Gemini CLI, built from OpenTelemetry logs and metrics.

Compare activity across tools, inspect individual sessions, and track token usage, estimated API-equivalent costs, latency, and errors.

Included views

  • Request counts, error rates, and request latency.
  • Input, output, cached, and reasoning/thinking tokens.
  • Estimated API-equivalent costs by model and over time.
  • Tool activity, model distribution, and session events.
  • Dedicated Claude Code, Codex CLI, and Gemini CLI sections.
  • Cross-agent logs for investigation.
  • Configurable Claude and Codex credit-reset countdowns.

Panel availability depends on the telemetry emitted by each CLI.

Requirements

  • Grafana with the VictoriaLogs data source plugin.
  • VictoriaLogs for telemetry logs and request events.
  • VictoriaMetrics configured through Grafana's Prometheus data source.
  • OpenTelemetry Collector receiving telemetry from the CLI tools.

The collector must normalize event fields to match the dashboard queries. Codex and Gemini cost panels also depend on collector-side cost enrichment.

The EUR conversion variable requires the usd_eur_exchange_rate metric.

Setup

  1. Enable OpenTelemetry telemetry in the CLI tools you use.
  2. Route their telemetry through the OpenTelemetry Collector.
  3. Export logs to VictoriaLogs and metrics to VictoriaMetrics.
  4. Map both data sources when importing the dashboard.
  5. Select the services to display using the Service filter.
  6. Adjust the reset-offset constants to match your account.

Cost estimates

Cost panels show estimated API-equivalent usage, not subscription charges or invoices.

Claude Code uses native cost estimates when available. Codex and Gemini use collector-side estimates for supported models. Accuracy depends on emitted token counts and configured pricing.

Credit-reset countdowns

Countdowns assume a repeating five-hour cycle. They do not query provider quota APIs.

For each account, calculate the offset as:

Unix timestamp of a known reset modulo 18000

The included offsets are examples and should be adjusted if your reset schedule differs.

Revisions
RevisionDescriptionCreated

Get this dashboard

Import the dashboard template

or

Download JSON

Datasource
Dependencies