AI agent time tracking Every AI agent you use, tracked in one place.
TimeUno tracks eleven AI coding agents without a plugin, from Claude Code to Ollama. This page lists what each one actually reports, and how agent-hours differ from the wall-clock time you were away.
TimeUno finds these agents on your computer on its own, no plugin or config file needed. The capture column below is specific to each agent: some report tokens and cost, some only report whether they are active.
Supported agents
| Agent | What TimeUno captures | Page |
|---|---|---|
| Claude Code | Working or waiting from the session transcript, the project folder, the model, tokens and cost | Docs |
| Pi and oh-my-pi | Working or waiting, the project folder, the model, tokens and cost | Docs |
| Codex CLI | Working or waiting, the project folder and the model | Docs |
| Cursor (editor and CLI) | Working or waiting from composer metadata or the CLI transcript, and the project folder | Docs |
| OpenCode | Session activity and the project directory, from its local database | Docs |
| Gemini CLI | Working or waiting, from its chat session files | Docs |
| GitHub Copilot (VS Code chat) | Working or waiting, from the chat transcript | Docs |
| Claude desktop app (Windows and macOS) | Working or waiting, from local session files when the app writes them | Docs |
| Windsurf | Whether it is active, from workspace state size, and the workspace folder | Docs |
| Aider | Whether it is active, from chat history file growth and length only | Docs |
| Ollama | Which models are loaded, from the local API | Docs |
Agent-hours versus wall-clock time
If you run three agents for an hour each, at the same time, that is three agent-hours of work, but only one wall-clock hour: TimeUno never doubles your day just because several agents ran inside it.
Wall-clock time is the union of every running session, the stretch where at least one agent was busy. Agent-hours is the sum of each session’s own working time, so parallel agents add past the hour they actually ran in. The Overview shows agent time beside your tracked hours, not folded into them: an hour in a meeting while Claude Code refactors in the background stays one hour away from your desk and one hour of agent work next to it, not two hours added together.
What never leaves your device
From every agent above, TimeUno sends metadata only: when a session starts and ends, whether the agent is working or waiting, the model, token count and cost where the tool reports them, and the project folder. Your prompts, the agent’s answers and your code never leave your device. See which AI tools TimeUno tracks for the full list and what each one exposes.
AI agents on a remote server
Claude Code, Cursor CLI, Codex CLI, Gemini CLI, OpenCode, Aider, Ollama, oh-my-pi and Pi can all run on a Linux server you reach over SSH, where the desktop app cannot see them. The headless agent tracks each one there, with its own project and whether it is working or waiting, on the same timeline as your laptop.
Updated 18 October 2026
Questions people ask
Which AI agents does TimeUno track?
Eleven, without a plugin: Claude Code, Codex CLI, Cursor (editor and CLI), OpenCode, Pi and oh-my-pi, Gemini CLI, GitHub Copilot, Windsurf, Aider, the Claude desktop app and Ollama. The table below lists what each one reports.
What does TimeUno send from each agent?
Metadata only: when a session starts and ends, whether the agent is working or waiting, the model, token count and cost where the tool reports them, and the project folder. Your prompts, the agent's answers and your code never leave your device.
If I run three agents at once, does that count as three hours of my day?
Not against your day. Three agents working for an hour each is three agent-hours, but TimeUno also shows the single wall-clock hour you were actually away, so your tracked time never doubles just because several agents ran together.
Does TimeUno track agents running on a server?
Yes, with the headless agent. It tracks Claude Code, Cursor CLI, Codex CLI, Gemini CLI, OpenCode, Aider, Ollama, oh-my-pi and Pi on a Linux server you SSH into, each running session on its own.