Skip to content

Apple has approved code signing for our macOS app, and we’re still securing it for Windows. In the meantime, follow the Windows guide to open Pip.

Agents

Three agents running. One is waiting for you.

Coding agents work while you look elsewhere. Pip lists every Claude Code and Codex session on your machine with what it is doing, how full its context is, and which one needs an answer.

macOS and WindowsPay once, updates included

The terminal is minimised. The agent stopped asking.

Several agents across several projects means several windows you are not looking at. Pip gives each session its folder name, its model, what it is doing right now, and how much of its context window is used — so the one that needs a reply is the one you notice.

Nothing here changes how your agents run. Pip reads Claude Code’s and Codex’s own files and the process table: no hooks, no edits to their configuration, nothing written under their data directories. Notifications stay quiet while you are looking at the Agents tab.

Working, idle, or waiting for your answer.Pip for Windows · Preview data
and OpenAI Codex CLI sessions
Claude Code
how full each session’s window is
Context
their files and processes are only read
No hooks

See what each agent is doing.

Rows say thinking, writing, compacted and continuing, idle, or waiting for you. A summary card counts the sessions working now and the ones that need an answer.

Watch the context window fill.

Each session shows how much of its context is used and where it compacts. Pip can notify you as a session approaches full, at a threshold you choose.

Get back to the right session.

Open the folder, bring forward the terminal or editor the agent runs in, or copy its resume command. End a session, or every idle one, when you are finished.

Pick up the session that is waiting.

  1. Run your agents as you already do.

    Start Claude Code or Codex in a project folder. Each session appears in Pip with its status and context, without any setup in the project.

  2. Check the list, or wait to be told.

    Scan the Agents tab for the sessions working and waiting, or turn on notifications for when one finishes, needs you, or nearly fills its context.

  3. Return, or wind it up.

    Jump to the agent’s app or folder, copy its resume command for later, or end sessions you no longer need — their conversations can still be resumed.

What Pip reads, and what it reports.

The Agents tab is built on files your agents already keep. These are the details worth knowing about where the readings come from and what the figures mean.

Which agents does Pip follow?
Claude Code and OpenAI’s Codex CLI. Pip reads Claude Code’s directory (~/.claude, or CLAUDE_CONFIG_DIR) and Codex’s (~/.codex, or CODEX_HOME), alongside the running processes behind them. Run either one once and the tab fills in. Pip does not start, prompt or steer an agent; it reports what yours are doing.
What is the API price figure?
It is what a session’s tokens would cost at the provider’s published API list prices, with the date of those prices shown beside it. It is an illustration of token use, not a bill: a Claude or ChatGPT subscription is not charged per token, and Claude Code’s own background calls are left out. Messages Pip has no price for are excluded and noted. You can switch the figure off in Settings.
Does Pip change how my agents work?
No. Pip only reads the session files each agent writes and the process table. It installs no hooks, edits no agent configuration, and writes nothing under their data directories. Ending a session is the one action that touches a running agent: it stops that process and everything it started, and tells you when something it spawned survived. The conversation itself can be resumed afterwards.

Keep every agent session in one list.

Every Pip tool. $29 USD, once. One machine at a time.