← Entitees AI
Entitees AI · Companion

A Development Dashboard with Companions

Works with Claude Code and Codex. One download reads both — whichever you have is picked up, and if you use both you get one combined view.

Your entity lives in a browser tab and reacts to your local sessions as they happen — thinking while the model thinks, busy while it runs tools, asleep when you step away. Underneath, a real dashboard of how you actually work.

Companion

Subscribe

$4.99/ month

Cancel anytime.

  • Works with Claude Code and Codex — one download, both CLIs, combined if you use both
  • The Bridge — every agent you have running, at once, as the crew of a ship
  • A live entity — any of the ten, reacting to your session in real time
  • Your usage dashboard — prompts, tool calls, tokens, streaks, top tools, projects and models
  • Private by design — reads the logs already on your machine; never your prompt text, never over the network
  • macOS, Windows, Linux — Python 3, nothing to install, delete the folder and it's gone
  • Download anytime while you're subscribed, including new versions
Subscribe — $4.99/month

Secure checkout by Stripe. You'll come straight back here to download.

Already subscribed

Download with your key

Use the download key you got after subscribing.

Lost your key? Reply to your Stripe receipt and we'll help.

A dashboard of how you actually work

Claude Code already writes a detailed log of every session on your machine. The Companion reads it and turns it into something you can look at: how much you prompted today, which tools ran, how many tokens went through which model, which projects took your attention, and how long your current streak is.

It updates while you work. Start a run in Claude Code and the numbers move within a second or two — you don't refresh anything.

What it shows

The entity is the point

A dashboard you have to go and look at is a dashboard you forget. So the numbers come with a face: whichever of the ten you pick sits in the corner and reacts to the same log — thinking while Claude is working, busy while it runs tools, pleased when a long run finishes, and asleep when you stop. You can tell the state of a run from across the room.

Private by design

It runs entirely on your own computer and makes no outbound network requests. It never reads the text of your prompts or Claude's replies — only timestamps, roles, token counts, model names, tool names and project folder names. The dashboard is served to 127.0.0.1, so nothing outside your machine can reach it. Nothing is written back to your Claude files, and deleting the folder removes it completely.

The Bridge — all your agents at once

Switch to Bridge and every agent you have running becomes a crew member on a ship, seen from above. Run six at once and you get six bots, each walking to the station that matches what it is doing right now: Bash to the engine room, Edit to fabrication, Read and Grep to the archive, browser and web tools to the sensor array, thinking-with-no-tool to the bridge itself. Claude Code and Codex agents stand side by side, and subagents get their own bot.

The Bridge view: a top-down spaceship interior with seven little robot crew members at different stations — the engine room, fabrication, archive, sensor array, ops and the bridge — around a glowing reactor core. live demo
Running live, with a staged crew. Seven agents aboard: the reactor glows with output tokens per minute, rooms light up only when someone is working in them, and each bot walks to the station matching its current tool. This is the same renderer the Companion ships — only the crew list is staged.

It is drawn from the same logs as the dashboard — timestamps, tool names, model names, folder names. Nothing extra is read, and nothing about the ship needs to know what anyone actually said. Included in the Companion; there is no second subscription.

What you need

Claude Code or Codex — either is enough — and Python 3, with no dependencies to install. Runs on macOS, Windows and Linux. $4.99/month, cancel any time, with downloads and updates for as long as the subscription is live.

Questions

Does it send my code or prompts anywhere?
No. The Companion reads the session logs Claude Code already writes on your machine, and it makes no outbound network requests at all. It never reads the text of your prompts or Claude's replies — only timestamps, message roles, token counts, model names, tool names and project folder names. The page it serves is bound to 127.0.0.1, so nothing outside your computer can reach it.
What exactly does it read?
The .jsonl session logs under ~/.claude/projects (Claude Code) and ~/.codex/sessions plus archived_sessions (Codex) — the ones those tools write themselves. Nothing is written back, and nothing is uploaded.
Do I need Claude Code for it to do anything?
You need Claude Code or Codex — either one. The Companion is built on the local session logs those tools already write, so it has nothing to show without at least one of them, and it combines both if you use both. It does not track Cursor, Copilot, the Claude or ChatGPT desktop chat apps, or the API directly: those keep conversations on the server, so there is no local log to read.
What do I need installed?
Python 3 and nothing else — no pip install, no dependencies. macOS usually has it via the Xcode command line tools; on Windows you can get it from python.org. Start it with start.command on macOS or start.bat on Windows. To remove it, delete the folder.
On a Mac, will it warn me about the download?
Yes, once — and it is expected. macOS says it “could not verify start.command is free of malware” because the file was downloaded and is not signed by a registered Apple developer. That is not a virus warning. Either open System Settings → Privacy & Security and click “Open Anyway”, or start it from Terminal with “sh start.sh”, which is not gated at all. The README covers both, and the launchers are a few lines of shell you can read first.
Which entities can I use in the Companion?
All ten: NYX, CLUTCH, Kodi, Zeta, DJ Catz, Dr. Eva 2.0, Dr. Eva, Mochi, NOVA and BIT. Switch between them any time; the Companion remembers your choice.
Does it need an API key?
No. The Companion watches your local Claude Code and Codex activity; it doesn't chat, so there is no key and no AI usage to pay for. The chat entities on the website are the ones that need your own Claude or ChatGPT key.
Can I cancel?
Yes, any time from the billing portal. You can download the Companion and any new versions for as long as the subscription is live.