Chrome extension
Know the moment
your AI job is done
Track progress and completion of AI agents and generative tools, in tabs and terminals. Local first, no account.
The Chrome Web Store listing is in review. Until it is published, the extension loads unpacked from the repository.
How it behaves
Four things it does, in motion
Everything below is drawn live in this page, with the extension's own state colours and icon geometry. Nothing here is a screen recording.
1. A job starts in a tab
On a site you switched on, a thin bar at the top of the tab fills as the job progresses. Hover the handle for the list of everything running on that page. Switch to another tab, minimise the window: the bar keeps working.
2. The toolbar icon carries the state
Blue while running, amber with a question mark the second an agent needs your answer, green when everything finished. Past two tasks the icon shows the count instead of a glyph, so one look is enough.
3. You are told once, when it matters
Finished, failed, stalled, or waiting for your answer. Not while you are looking at the tab, and not for a job that took three seconds. Per tool you choose which of those are worth an interruption, and whether they also go to a messenger.
4. Terminal agents, not just tabs
Claude Code, Gemini CLI and Codex report every step through a relay you run yourself on 127.0.0.1. The case that saves the most time is the last line: the agent is asking for permission while you are somewhere else, and the tray turns amber.
Icon legend
Coverage
What it watches
Sites in a tab
Sora, Runway, Kling, Midjourney, Suno, Luma, HeyGen, ElevenLabs, and the research modes of Claude, ChatGPT, Gemini and Perplexity. Six detection layers read the page, from a real progress element down to a spinner turning into a video.
Terminal agents
Claude Code, Gemini CLI, Codex CLI and Cursor, through hooks and a relay that stays on your machine. One line in the agent's own config, one Node process, no packages.
Local pipelines
ComfyUI and Ollama, plus anything else you write: the task format is a small open protocol, so a CI job or a script of your own reports into the same tray.
Permissions
What it will not do
- No site access until you grant it. The package ships with no host permissions at all. Each site you switch on is a separate Chrome prompt, and turning it off removes the access completely.
- No account and no server of ours. There is nothing to sign into. A fresh install makes no network request whatsoever.
- No analytics, no telemetry, no remote code. Everything that runs is in the package you installed. Nothing is fetched and executed.
- Every outbound request is listed. Settings, Data and privacy, holds a live log of them, next to one switch that blocks all of them.
Inside
The parts you configure
Trust
Check it instead of believing it
Every claim above is verifiable without taking anyone's word.
- Reproducible build. A release tag carries the SHA-256 of the exact zip that was uploaded to the store, so the code you read is the code you installed.
- The permission screen is the proof. Chrome shows what an extension may do at install time. This one asks for nothing until you enable a site.
- Its own audit log. The extension records every request it has ever made and shows it to you in its settings.
Questions: hello@example.by or the issue tracker.