PiDr reads your repo, edits it, runs the shell, drives a browser and the desktop, and keeps what it learned for the next session. One process, macOS, Linux and Windows — no WSL, no bridge, no second tool to keep in sync.
Not a picture of it. This is PiDr itself, running on the server — click into it and type. Every visitor gets their own session.
The tools are native to the process. ripgrep, glob and the shell run inside it, so nothing depends on what happens to be installed on the machine.
Hashline points at the lines that change. Edit a stale file and the anchors diverge — the patch is rejected before it corrupts anything.
The shell keeps its session across calls, with 58 utilities ported in-process. No fork, no exec, no missing binary on Windows.
Fan work out, isolate it in its own workspace, and pull a field back out of a subagent's output by path.
It writes facts mid-run, captures reusable lessons, and loads a model of the repo on the first turn of the next session.
/review sweeps a branch or a commit and ranks every issue P0 to P3, with a confidence score and a ship-or-not verdict.
It drives a real browser and the host itself — windows, screenshots, native input, the accessibility tree. One of the tools, not the product.
curl -fsSL https://pidr.ink/install | shmacOS · Linuxirm https://pidr.ink/install.ps1 | iexWindows · PowerShell