AI Observatory
GitHub Demo

Setup

One command, about a minute

You have seen the demo. This puts the same dashboard on your machine, with your own numbers in it. Two ways to get there — pick whichever you are already sitting in front of.

Easiest — no terminal

Let your coding agent do it

Paste this into Claude Code, Codex, Cursor or Gemini CLI. It installs everything, then tells you what it found in your own usage.

Clone https://github.com/jxxyx-bloop/ai-observatory, then from its observatory/ folder run: python3 observe.py setup — it checks the machine, collects my local transcripts, builds the dashboard and installs a launcher. Then run python3 observe.py insights and summarise the top three findings for me.

It only reads transcript files already on your disk and writes inside the project folder. Nothing is uploaded, nothing is deleted, and collecting costs no tokens.

Prefer the terminal?

Paste one line yourself

git clone https://github.com/jxxyx-bloop/ai-observatory.git && cd ai-observatory/observatory && python3 observe.py setup

What that one line does

  1. Checks your machine

    Confirms Python 3.9 or newer and finds which AI coding tools have run here. There are no packages to install — the whole engine is standard library, which is why this is one command and not a dependency file.

  2. Updates itself

    Fast-forwards to the latest version if it can reach GitHub. Never a merge and never a rebase — no network, or local edits of your own, simply means it keeps the version you have.

  3. Reads your transcripts the slow part

    Your coding agents already write a log of every turn. This reads those files, counts them, and throws the text away. Read-only, no network, and it costs no tokens.

  4. Builds your dashboard

    Prices every turn against a local rate card and runs fifteen checks over the result, so the page opens on findings rather than on numbers you still have to interpret.

  5. Puts it in your Dock, and opens it

    Creates the app, pins it, and schedules a quiet refresh for 09:00 each morning and for every login. Then your dashboard opens — which is how you know it worked.

Your terminal will look like this

AI Observatory — setting up

1/5  Checking your machine
      ✓ Python 3.12
      ✓ Nothing to install — standard library only
      ✓ 3 provider directory(ies) found.

2/5  Updating to the latest version
      ✓ already up to date

3/5  Reading the transcripts already on this disk
      nothing is uploaded, and no tokens are spent
      ✓ 18,095 new events from 274 sources

4/5  Building your dashboard
      ✓ 18,095 turns, 311 sessions, 9 findings

5/5  Putting it in your Dock
      ✓ launcher ~/Applications/AI Observatory.app
      ✓ pinned

Done. Opening your dashboard now.

Seeing 0 new events from 0 sources? That is not a failure — no supported tool has run on this machine yet. It fills in sample data so you still finish with a dashboard, and the page says so at the top until your own usage arrives.

Not on a Mac? The same command writes Open AI Observatory.cmd (Windows) or a .sh script (Linux) beside the project. Double-click it, or pin it to your taskbar.

That's it — here's what happens now

Tomorrow, and after you restart your Mac
Click the icon in your Dock. It refreshes first, then opens — so what you see is always current. Nothing needs to be running in the background for this to work.
Each morning at 09:00, and again at login
It collects quietly and rebuilds the dashboard, without opening a window or interrupting you. If your Mac is asleep at nine it runs when you open it, and if it was shut down it runs when you log back in — so what you see is current whenever you next look, not only on the days the machine happened to be awake.
If you ever lose it
Spotlight (⌘ Space) and type Observatory. It also lives in ~/Applications.
If a number looks stale
The dashboard dates itself — a report more than a day old says so above the title and hands you the one line that refreshes it. You never have to guess whether you are looking at old numbers.
To undo all of it
Run python3 observe.py install --remove. It removes the app and the schedule, and never touches your collected data.

If something did not work

Five things go wrong, and all five have a one-line answer. python3 observe.py doctor checks every step and prints the fix for whichever failed.

“python3: command not found”
macOS and Linux ship Python 3, so this is almost always Windows. Install it from python.org, tick Add Python to PATH, then use python instead of python3.
“sync: 0 new events from 0 sources”
No supported tool has run on this machine yet, or its transcripts live somewhere non-standard. Expected on a fresh laptop — the sample data from step 1 still shows the whole product.
The page says there is no digest yet
Collecting and rendering are separate steps. Run python3 observe.py all, which does every step in order.
Nothing opened in my browser
The file is still there — open dist/observatory.html from the project folder. Scheduled runs never open a browser on purpose, so a morning refresh cannot steal focus while you work.
macOS says the app is from an unidentified developer
That warning belongs to downloaded apps. This one is generated on your own machine, so it should never appear — if it does, the app was copied from another computer. Delete it and run python3 observe.py install again.