lpm
AGENTS.md150 starsChanged 5 days agoMITView on GitHub
lpm starts, stops, duplicates, and switches between local dev projects, with a built-in terminal for running AI coding agents alongside services. The desktop app runs on macOS, Linux and Windows. - desktop/frontend/ — Tauri 2 desktop app: React/TypeScript UI in src/, Rust backend in src-tauri/ - cli/ — Rust CLI (lpm) - mobile/ — lpm Link, the phone companion apps: - PROTOCOL.md — the wire contract both apps implement - shared/web/ — xterm.js terminal and markdown pages, bundled into both…
- Installs packages
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository22 other files it gives its agents.
- mobile/AGENTS.mdAGENTS.md
- website/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- mobile/CLAUDE.mdCLAUDE.md
- website/CLAUDE.mdCLAUDE.md
- android-cliSkill
- app-review-videoSkill
- app-store-releaseSkill
- app-store-screenshotsSkill
- edge-to-edgeSkill
- lesson-publishSkill
- lessonSkill
- navigation-3Skill
- short-dailySkill
- short-gameSkill
- short-publishSkill
- short-raceSkill
- shortSkill
- lpm-cliSkill
- lpm-configSkill
- lpm-memorySkill
- lpmSkill
The file
We split this AGENTS.md into the rules it gives an agent
7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 45
Use the shared seams instead of raw platform APIs: ipc (socket types), osproc (kill/alive/detach, and osproc::command so console children …
- Very specific: line 52
Check every Rust change for Windows with AWS_LC_SYS_PREBUILT_NASM=1 cargo check --target x86_64-pc-windows-gnu (needs brew install …
- Distinctive: line 58
Keep files focused: don't grow a file past ~400 lines — put new features in their own module
- Very specific: line 60
When a lpm desktop app change touches CLI-supported functionality, update the lpm CLI and the lpm-cli skill in the same change.
- Very specific: line 61
Keep lpm Link (mobile/) in step with the desktop app: when a desktop change touches something the phone also offers (actions, services, …
- Distinctive: line 72
Keep connected computers (+ above your projects → Another computer or Server) in step with the desktop app: a project on a paired machine …
- Distinctive: line 76
Don't commit or push unless asked
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Agent Guide — lpm
lpm starts, stops, duplicates, and switches between local dev projects, with a built-in terminal for running AI coding agents alongside services. The desktop app runs on macOS, Linux and Windows.
Layout
desktop/frontend/— Tauri 2 desktop app: React/TypeScript UI insrc/, Rust backend insrc-tauri/cli/— Rust CLI (lpm)mobile/— lpm Link, the phone companion apps:PROTOCOL.md— the wire contract both apps implementshared/web/— xterm.js terminal and markdown pages, bundled into both apps' WebViewsios/— iOS app (SwiftUI);.xcodeprojis generated — runxcodegen generateinmobile/iosafter adding filesandroid/— Android app (Kotlin, Compose, Hilt, Navigation 3; modules built from thebuild-logicconvention plugins). Inmobile/androidrun./gradlew spotlessApply, then./gradlew check assembleRelease. It's a port of the iOS app on Google's architecture recommendations: readmobile/android/PORTING.mdfirstparity/— the same Maestro flows on both apps in an iPhone simulator and an Android emulator, with a side-by-side report (node mobile/parity/parity.js all --open)
website/— Next.js marketing site (lpm.cx). Every interactive demo on it is the desktop UI itself in an iframe:desktop/frontend/src/demoanswers its commands from sample data in memory and plays each page's tour against the real UI (npm run dev:demoindesktop/frontend,?seed=samplefor the sample workspace), built intowebsite/public/demo-app./changelogshows each GitHub release's notes, whichscripts/release.shhas Claude write into the tag;website/lib/changelog.jsonholds older releases' notesrelay/— Cloudflare Worker that lets a browser join a shared tab: a Durable Object room per share, forwarding Noise-encrypted messages, and the guest page (lpm-share-relay.lpm-share-relay.workers.dev).npm run dev/npm test/npm run deploy; see its README. It's for tab sharing only: never give it other features' routes or storagetailnet/— built-in Tailscale (Go, tsnet):cmd/lpm-tailnetis the desktop sidecar (scripts/build-tailnet.mjs),ios/the static library lpm Link links (mobile/ios/build-tailnet.sh). Needs Go (brew install go)
Conventions
- Clean code and best practices; no comments unless the reasoning isn't clear from the code itself
- The desktop app targets macOS, Linux and Windows, and Linux is also a
headless host (the same binary on a server, driven from a Mac). macOS
behaviour must not change when another platform is added: gate platform code
with
#[cfg(target_os = "macos")]/#[cfg(windows)]/#[cfg(unix)]and give every platform a working path or a clear error — never an unconditionalscutil/lsof/pbcopy/open/osascript/Keychain call,std::os::uniximport orlibccall - Use the shared seams instead of raw platform APIs:
ipc(socket types),osproc(kill/alive/detach, andosproc::commandso console children don't flash a window on Windows),fsperm(mode bits, locks),sys(PATH,which, login shell,headless()), andsrc/platform.tsin the frontend - Windows runs shell lines through Git for Windows bash (
sys::login_shell) - Shortcuts: macOS uses ⌘; Linux/Windows use Ctrl+Shift for app chords and leave plain Ctrl+letter to the terminal and composer. Show platform-correct labels
- Check every Rust change for Windows with
AWS_LC_SYS_PREBUILT_NASM=1 cargo check --target x86_64-pc-windows-gnu(needsbrew install mingw-w64, the rustup target, a placeholderbinaries/lpm-cli-x86_64-pc-windows-gnu.exe, andnode scripts/build-tailnet.mjs x86_64-pc-windows-gnu) - One React component per file
- Keep files focused: don't grow a file past ~400 lines — put new features in their own module
- Bump
versionincli/Cargo.tomlon anycli/change — patch for fixes, minor for new commands/flags — so stale installed binaries stay detectable vialpm --version - When a lpm desktop app change touches CLI-supported functionality, update the lpm CLI and the
lpm-cliskill in the same change. - Keep lpm Link (
mobile/) in step with the desktop app: when a desktop change touches something the phone also offers (actions, services, terminals, automations, Git, notes, memory, project config), update the mobile app in the same change so it looks and behaves the same — same fields, defaults, placement and colors. Data the phone needs goes throughremote.rs,mobile/PROTOCOL.mdand Demo Mode's seed (mobile/ios/Sources/LpmMobile/Demo*.swift, mirrored inmobile/android/core/demo— regenerate its goldens withcore/demo/ios-parity/generate.sh), and a phone edit writes the config exactly as a desktop edit would. Once the Android app implements a feature, change it on both phones in the same change, and keep its parity flow inmobile/parity/flowspassing on both - Keep connected computers (+ above your projects → Another computer or Server) in step with the desktop app: a project on a paired machine must work the same as a local one
- Give every new Tauri command a demo answer: handle it in a
desktop/frontend/src/demomodule, or list the empty value of its return type insrc/demo/defaults.ts(defaults.test.tsfails until you do) - Don't commit or push unless asked
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

