More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    JH AI Observatory — Claude Code & Codex Usage Tracker

    1

    Local-only dashboard for your Claude Code and Codex usage: tokens, conversations, models, projects and API-equivalent value. Your logs never leave your machine.

    $1

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    JH AI Observatory — Claude Code & Codex Usage Tracker

    JH AI Observatory — Claude Code & Codex Usage Tracker

    Example session with this skill installed

    Show me my AI coding usage dashboard.

    • Read your context and instructions
    • Compiled the jh ai observatory

    JH AI Observatory is installed and running locally.

    Setup summary

    StepResult
    InstallCloned to ~/jh-ai-observatory, Node.js 18+ found, no npm dependencies
    ServerRunning at http://127.0.0.1:5858 (local only)
    Claude Code logs~/.claude/projects/**/*.jsonl
    Codex logs~/.codex/sessions/**/rollout-*.jsonl
    Data sharingNone. Logs are read and stored on your machine only

    I've opened the dashboard in your browser. Because this is a fresh install, it doesn't know which projects to track yet, so it's showing the

    Choose your project directories card.

    Next steps

    1. Click Add directories and add the folder that holds your projects, for example ~/Developer.
    2. Make sure it's checked, then click Save & collect.
    3. Your usage appears right away: turns, conversations, tokens by model and project, and API-equivalent value at list prices (a reference, not your bill).

    After that, the dashboard collects new usage every 5 minutes. Want to try it with demo data first? Ask me to run it with --mock on port 5656.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    See how you actually use AI coding agents. This skill installs and launches JH AI Observatory, a local dashboard that reads the logs Claude Code and Codex already write to your machine and turns them into clear usage views by model, project and day.

    What it does

    • Measured, not guessed: token counts come straight from the logs. Missing data shows as — and unknown models as Unpriced, never 0.
    • Usage breakdowns: turns, conversations, input, output and cache tokens, by model, source and project.
    • API-equivalent value: prices usage at list API rates as a reference point. It's not your bill.
    • Incremental collection: only new log data is read, refreshed every 5 minutes or on demand.
    • Local only: binds to 127.0.0.1, with no telemetry and no CDN.

    How it works

    1. The agent runs scripts/install.sh, which clones the open-source repo and checks for Node.js 18+.
    2. It runs scripts/run.sh and gives you the URL: http://127.0.0.1:5858.
    3. On first launch, you add your project folders in Add directories and click Save & collect.

    Frameworks & tools

    Node.js 18+ with zero npm dependencies, vanilla JavaScript and hand-rolled SVG charts. Reads Claude Code (~/.claude/projects) and Codex (~/.codex/sessions) logs.

    Why this beats prompting it yourself

    Asking an agent to total its own usage means re-parsing raw .jsonl logs every time, with no history, no charts and easy mistakes around cache tokens. This gives you a persistent, tested dashboard (60 automated tests) that updates itself.

    Use cases

    • See which models and projects use the most tokens
    • Compare Claude Code and Codex usage side by side
    • Check how much of your input is served from cache
    • Track daily, weekly or monthly usage trends

    Privacy

    Your logs never leave your machine. The only optional exception is AI insights: if you add your own Anthropic API key, an aggregated summary (not raw logs) is sent to generate the insights.

    Known limitations

    • Requires Node.js 18+ and Git.
    • The dashboard stays empty until you add project directories in Settings.
    • API-equivalent value is a reference at list prices, not your actual bill.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    Requires Node.js 18+ and Git. Empty until user adds project directories in Settings. Only tracks Claude Code and Codex logs. API-equivalent value is a list-price reference, not your actual bill.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 day ago

    Accounts
    GitHub
    Software
    GitNode.js 18+
    Claude Code
    orCodex CLI
    Optional
    Anthropic API key (for AI insights)

    Works on macOS, Linux and Windows. On Windows, the install scripts run through Git Bash or WSL (Claude Code uses Git Bash by default).

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables

    Allowed Hosts

    github.com

    File Scopes

    ~/.claude/projects/**
    ~/.codex/sessions/**
    ~/jh-ai-observatory/**

    Network access is only used to clone the open-source repo from GitHub at install (and git pull to update). The dashboard reads local Claude Code and Codex logs, writes only inside its install folder, and serves on 127.0.0.1. No data is sent anywhere unless the user adds their own Anthropic API key for optional AI insights.

    Listed1 day ago

    What's inside

    Frequently Asked Questions

    It turns the logs Claude Code and Codex already write into a local dashboard: turns, conversations and tokens by model, project and day, plus an API-equivalent value at list prices. It's a reference, not your bill.

    No. Logs are read and stored on your machine, and the dashboard only listens on 127.0.0.1. The one optional exception is AI insights: if you add your own Anthropic API key, an aggregated summary (not raw logs) is sent to generate them.

    The skill runs in any SKILL.md-compatible agent, but the dashboard only reads Claude Code and Codex logs for now.

    The skill (SKILL.md plus install and run scripts) that sets up and launches the dashboard for you from a single request. The dashboard is downloaded from GitHub during install.

    Node.js 18+, Git, and Claude Code or Codex CLI. There are no npm dependencies. It works on macOS, Linux and Windows; on Windows, the scripts run through Git Bash or WSL.

    New in LLM Ops