Federico Jaramillo Martinez
69b125b001
A session spawned into another worktree recorded a parent that no
listing contained, so the row showed only "parent unavailable" and its
parent's row looked childless. Both facts were accurate and useless:
neither said where the related session actually was.
Report both directions from the session store instead. A missing
parent's cwd and id come from its own file header, so one 4 KB read per
distinct missing parent resolves it without listing other workspaces;
children are counted by listing sibling workspaces and matching the
parent path they already recorded, needing no header reads. Reads are
memoized per path because Pi writes headers once, and the cache is
released on dispose. Both directions are best-effort: an unreadable
header or an unlistable worktree leaves a session unannotated rather
than failing the listing.
In the browser, an orphan child keeps the same child marker as a nested
one, dimmed, so it no longer renders as a root; whereabouts are stated
once on the meta line ("parent in feature/foo", "2 children
elsewhere"), where a clamped title cannot hide them. A "Go to parent
session" action switches to the owning workspace and selects the
parent. Live session.created events keep child counts current instead
of leaving them stale until the next listing.
Session and workspace paths reach the browser from two producers: store
enumeration for a listing, and the live runtime for a broadcast. They
are now compared through one normalizing helper, so tree nesting and
child counts cannot silently miss a link when only a trailing separator
differs.
Extract the shared "workspaces of the project containing this cwd"
lookup out of ProjectScopedSpawnTargetResolver so spawn targeting and
child counting share one implementation, and register it regardless of
whether spawning is enabled: children can predate a config change, and
the tree should stay honest about them either way.
PI WEB
PI WEB is a web UI for Pi Coding Agent that keeps agent sessions running in real workspaces on your machine or server.
Run agents where your code, tools, credentials, and build caches live. Supervise them from any browser.
Website and docs: https://pi-web.dev/
Why PI WEB?
Agentic development works better when the work environment is persistent.
PI WEB lets you:
- keep Pi Coding Agent sessions alive after browser disconnects;
- run agents inside real repositories and git worktrees;
- supervise multiple sessions in parallel;
- switch between laptop, phone, tablet, and desktop;
- use a server, workstation, or remote dev box as your agent runtime;
- manage projects, workspaces, files, terminals, sessions, and remote machines from one web UI.
Your browser is the control surface. The work stays where it can keep running.
Quick start
Requirements:
- Node.js 22.19.0 or newer
- npm
- Pi Coding Agent
>=0.82.1 <0.83, configured for your user - git and the development tools your agents need
Install and start PI WEB as per-user services:
npm install -g @jmfederico/pi-web --allow-scripts=node-pty
pi-web install
pi-web doctor
On npm 12, the scoped flag lets node-pty prepare its required native module without enabling install scripts for other dependencies.
Then open:
http://127.0.0.1:8504
Useful commands:
pi-web status
pi-web logs
pi-web restart
pi-web doctor
pi-web version
pi-web uninstall
For more install options, including one-line install, Pi package install, WSL/manual usage, and remote access, see the installation guide.
Core model
PI WEB organizes work like this:
Machine a local or remote PI WEB runtime endpoint
Project a folder on that machine
Workspace a git worktree, or the project folder for non-git projects
Session a Pi Coding Agent chat running inside a workspace
A typical flow:
- Add a project.
- Choose a workspace or git worktree.
- Start a session.
- Let the agent work.
- Come back later from any browser.
Remote-first development
PI WEB is designed for remote AI-driven development.
Instead of tying agent work to your laptop session, run PI WEB on a machine that stays available: a server, desktop, cloud VM, home lab machine, or remote dev box.
Use a private network, SSH tunnel, trusted reverse proxy, or federated PI WEB machine setup when accessing it remotely.
Read more: Remote-first development
Machines and fleets
PI WEB can register other PI WEB runtimes as remote machines. One browser-facing PI WEB instance can proxy projects, files, git state, sessions, terminals, activity, Pi package management, and selected-machine settings from trusted remote machines.
When a remote machine is selected, Settings tabs label their target. Pi packages, PI WEB plugin enablement, session daemon toggles, external file access, and upload defaults target the selected machine. Gateway/server settings such as host, port, allowed hosts, registered machines/tokens, and keyboard shortcuts stay local to the gateway/browser.
Read more: Fleet and machines guide
PI WEB plugins
PI WEB supports trusted browser-side plugins that can add actions, workspace panels, and workspace metadata. Use Settings → PI WEB plugins to enable or disable them on the selected machine.
Pi packages are a separate Pi package-manager concept. A Pi package may include a PI WEB browser plugin, but installing a package and enabling its browser plugin are different operations.
Read more: PI WEB plugin guide and API
Configuration
Global config lives at:
$PI_WEB_CONFIG
~/.config/pi-web/config.json
Project-local PI WEB config lives at:
<project>/.pi-web/config.json
Common configuration includes host/port, path access, uploads, PI WEB plugin enablement, shortcuts, and session daemon options. In Settings, machine-affecting config targets the selected machine; gateway host/port/allowed-hosts, remote machine registration, tokens, and keyboard shortcuts stay local.
Read more: Configuration reference
Development
Clone the repository and run:
npm install
npm run dev
Open the Vite URL, usually:
http://localhost:8505
For the split development setup:
npm run dev:sessiond
npm run dev:web
npm run dev:client
Validate changes with:
npm run verify
Security model
PI WEB assumes trusted users, trusted repositories, and trusted server paths.
It is not a sandbox, permission system, or multi-tenant platform. Do not expose it directly to the public internet without a trusted network, firewall, VPN, SSH tunnel, or authenticated reverse proxy.
Documentation
License
MIT © 2026 Federico Jaramillo Martinez. See LICENSE.