How do I monitor Gemini CLI sessions?

Install ClawMetry (pip install clawmetry) and run it. The daemon finds Gemini CLI on the machine on its own and renders every session, turn, tool call, model, and token count in one dashboard. Gemini CLI records its chats without being asked, so the history you already have shows up on the first run. Nothing to configure in Gemini CLI, and no SDK to add.

Do Gemini CLI subagents show up separately?

Yes. Gemini CLI can hand work to specialist subagents, each with its own context window and its own tools. ClawMetry shows each one as its own session nested under the parent that launched it, with the brief it was handed, the answer it sent back, its own tool calls, and its own tokens. A subagent that quietly burned half the turn is visible instead of being folded into the parent total.

How accurate is the cost figure for Gemini CLI?

Token counts are exact and so is the model, because Gemini CLI records both per turn: input, output, cached prefix, and the reasoning tokens. Dollars are never recorded by any agent CLI, so ClawMetry derives them from those exact tokens at that model's published rates, and it prices the cached prefix at cache rates rather than charging it as fresh input. If you signed in with a Google account, your requests come out of a daily request quota instead of a bill, so read the dollar figure as what that work is worth at API rates.

Does this page cover Antigravity and the Gemini Code Assist IDE extensions?

This page is the Gemini CLI, the agent in your terminal. Antigravity, Google's agent-first IDE, is a separate runtime in ClawMetry with its own page at /runtimes/antigravity. The Gemini Code Assist extensions for VS Code and JetBrains are neither: they are an in-editor assistant, not a terminal agent, and they are not claimed here.

Is ClawMetry safe to point at a running Gemini CLI?

Yes. ClawMetry only observes. It never writes to Gemini CLI, never sends a prompt on your behalf, and a session it cannot read is skipped rather than mis-reported. A session you deleted stays deleted.

How much does ClawMetry cost for Gemini CLI?

Free forever on your own machine. Cloud starts at $9/node/mo (Starter) or $19/node/mo (Pro).

Gemini CLI × ClawMetry

Observability and cost tracking for Gemini CLI

See every Gemini CLI session, turn, tool call, subagent, and token as it happens. ClawMetry finds Gemini CLI on your machine on its own and renders sessions, transcripts, token splits, and cost in one dashboard, next to every other agent you run.

Supported runtime Zero config, auto-detected Read-only Free trial Official Gemini CLI site ↗

📦 631k+ installs 🌍 128+ countries ⭐ 392 on GitHub 🏆 #5 Product Hunt

What ClawMetry sees on Gemini CLI

Gemini CLI is Google's open-source terminal agent, the one you start with npx @google/gemini-cli, npm install -g @google/gemini-cli or brew install gemini-cli. It is Apache-2.0, it has over 106,000 GitHub stars, and it runs Gemini models against your repo with file edits, shell commands, web fetch, Google Search grounding, MCP servers, extensions, and specialist subagents. Once ClawMetry is running, your Gemini CLI activity shows up as:

Sessions per project: one row per Gemini CLI session, labelled with the directory you ran it in, when it started, when it last moved, message count, and its token totals.
The full transcript: your prompts, the model's answers, its reasoning, and every tool call with its arguments, its result, and the status it finished in, in the order they happened.
Subagents as their own sessions: Gemini CLI can delegate to specialist agents that each get their own context window and their own tools. Each one nests under the turn that launched it, with the brief it was handed, what it sent back, its own tool calls, and its own tokens.
Token splits that include the cache: fresh input, the cached prefix, output, and the reasoning tokens Gemini spends before answering, counted per turn and added up per session. A long agentic turn is not reported as one call.
Cost from the real model: Gemini CLI records which model produced each turn, so cost is derived from exact tokens at that model's rates, with the cached prefix priced as cache and not as fresh input. Nobody's CLI writes dollars to disk, so the figure is derived, never invented, and never doubled.
Tool-using turns counted once: a turn that calls a tool is recorded again once the call resolves. ClawMetry keeps the final version, so those turns are not billed twice.
Next to everything else: Gemini CLI sits in the same dashboard as Claude Code, Codex, Cursor and the rest, so fleet cost per hour is one number instead of six tabs. Qwen Code, which began as a fork of Gemini CLI, stays a separate runtime here, so the two never merge into one blurred total.

What this covers, exactly: the Gemini CLI, the agent in your terminal. Two neighbours are separate things and are not silently folded in:

Antigravity, Google's agent-first IDE, is its own runtime in ClawMetry with its own page at /runtimes/antigravity.
Gemini Code Assist in VS Code and JetBrains is an in-editor assistant rather than a terminal agent, and nothing on this page claims it.

Read-only by design: ClawMetry never writes to Gemini CLI and never sends a prompt on your behalf. It reads what Gemini CLI already recorded, after the fact. A session you deleted stays deleted.

Enable it in about two minutes

1. Install ClawMetry

Pick the line that matches your OS. Windows users: paste the install.cmd line, not the shell one.

# macOS or Linux
curl -fsSL https://clawmetry.com/install.sh | bash

# Windows (cmd.exe or PowerShell)
curl -fsSL https://clawmetry.com/install.cmd -o install.cmd && install.cmd

2. Run the daemon

clawmetry

The daemon finds Gemini CLI on the machine and starts reading immediately. No config, no flag to turn recording on, and no need to restart Gemini CLI.

3. Open the dashboard

Open http://localhost:8900. Switch the runtime chip to Gemini CLI.

Start the free trial →

The one thing this catches that Gemini CLI alone does not

Inside the CLI, /stats tells you what the session you are sitting in has used, cached tokens included. What it cannot tell you is the shape of the week: how many sessions you left behind in how many repos, which ones burned tokens re-reading the same files, how much of yesterday went to subagents you never watched run, and how Gemini CLI's total compares with the Claude Code and Codex runs happening beside it. On a Google sign-in the meter that actually runs out is a daily request quota, not a bill, and nothing in the terminal shows you how fast you are spending it across projects. ClawMetry keeps that view: spend per hour across every session and every runtime, subagents billed separately instead of hidden inside a parent, repeated-call detection, and token-growth curves per session. Set a threshold and the approval queue pauses before the next call. Kill switch (Pro) stops it.

Starred by the people building AI

Engineers and founders from OpenAI, Google, PostHog and more have starred ClawMetry on GitHub. One of them helped build it.

Peter SteinbergerOpenAI

Creator of OpenClaw and founder of PSPDFKit, now at OpenAI.

Matt Van HornJune · Weber

Co-founder of June, the smart oven acquired by Weber. ClawMetry contributor.

Alex KrentselGoogle

Systems research engineer at Google and PhD student at UC Berkeley.

Manoel Aranda NetoPostHog

Software engineer at PostHog, previously mobile SDKs at Sentry.

Kody KendallLlamaPress AI

Co-founder of LlamaPress AI, an open-source AI web-app builder.

All 26 runtimes ClawMetry observes

Gemini CLI is one of 26. Free on every plan: OpenClaw, NVIDIA NemoClaw, Goose. Paid on Cloud Pro or Self-Hosted Pro: the rest.

OpenClaw Claude Code Cursor GitHub Copilot Codex Cline OpenHands OpenCode NVIDIA NemoClaw Aider Goose n8n Qwen Devin Antigravity Grok Hermes Pi Deep Agents Kimi CLI DeepSeek Harness Exo NanoClaw PicoClaw QM Gemini CLI (this page)