LaunchPad

See, diagnose, and control macOS launchd jobs without memorizing launchctl.

_{Rendered from the current Slint interface with real, non-sensitive user LaunchAgents on macOS. The home path is abbreviated.}

What I engineered

1. A native Rust GUI over launchctl discovery, status, diagnostics, and job actions.
2. A layered architecture that separates filesystem and process adapters from domain rules, services, and UI state.
3. Explicit safety gates for privileged and destructive operations, including read-only system daemons and two-step deletion.

Rust + Slint · 66 passing tests · explicit capability checks before actions

Why LaunchPad exists

macOS background jobs are powerful, but their state is split across plist files, launchctl commands, unified logs, permissions, and multiple scopes. The result is a workflow that is easy to inspect incorrectly and risky to edit casually.

LaunchPad brings those surfaces into one reviewable workspace:

Outcome	Current implementation
Find jobs	Scans user agents, global agents, and system daemons with search, scope, status, and attribute filters
Diagnose failures	Combines fast launchctl list status with on-demand launchctl print details and recent unified logs
Control safely	Starts, stops, kickstarts, loads, unloads, enables, and disables only when scope and permissions allow
Edit with review	Builds plist XML from a standard editor, runs diagnostics, and requires confirmation before destructive changes

Safety is part of the product

LaunchPad does not treat every discovered job as equally mutable.

flowchart LR
    A[Discover job] --> B{Scope and permission check}
    B -->|System daemon| C[Read-only inspection]
    B -->|Controllable agent| D[Enable safe actions]
    D --> E{Destructive action?}
    E -->|No| F[Execute through launchctl adapter]
    E -->|Yes| G[Review target and confirm]
    G --> F

Loading

The same capability model is used by the UI and service layer, so disabled controls are backed by domain rules rather than presentation alone.

Current status

LaunchPad is a source prototype, not a signed or notarized macOS release yet.

Implemented in the current source build

1. Job discovery across ~/Library/LaunchAgents, /Library/LaunchAgents, and /Library/LaunchDaemons.
2. Search, scope, status, starred-only, and plist-attribute filters.
3. Bulk status plus on-demand runtime details.
4. Guarded trigger actions and two-step deletion.
5. Persistent starred jobs and copyable diagnostics.
6. Standard plist creation and editing with XML preview and rule-based diagnostics.
7. History and short live-capture log views.
8. A command palette over the same service actions.

Experimental

1. AI-assisted plist suggestions and provider adapters.
2. QuickLaunch snapshot, helper, and menu-bar integrations.
3. Advanced or undocumented plist keys represented as expert entries.

Not shipped yet

1. A signed, notarized, downloadable app bundle.
2. Automated end-to-end tests against real macOS launchctl state.
3. A fully general editor for arbitrary nested plist structures.

See the macOS manual test plan for the current main-path checklist.

Architecture

Slint UI
   │
Application controller
   │
Service layer           job · action · delete · plist · logs · AI · QuickLaunch
   │
Domain rules            status · capabilities · filters · plist model
   │
Adapters                launchctl · filesystem · unified logs · clipboard · providers

Two choices keep the system responsive and testable:

1. Refresh uses one bulk launchctl list query; deeper launchctl print work happens only for the selected job.
2. Process execution and filesystem access sit behind traits, allowing service behavior to be tested with mocks.

Read the full architecture overview.

Build from source

Requirements

1. macOS with a GUI session.
2. Rust stable. The current dependency graph is validated with Rust 1.94 or newer.

git clone https://github.com/Jacky040124/launchd-gui.git
cd launchd-gui
cargo run --bin launchpad

LaunchPad reads the normal user and system launchd locations. Test write actions on a disposable user LaunchAgent first.

Validate

cargo fmt --all -- --check
cargo clippy --all-targets --all-features -- -D warnings
cargo test --all-targets

The repository includes 66 passing unit and integration tests across domain logic, adapters, job actions, deletion, starring, and QuickLaunch behavior. Real system behavior still requires the macOS manual test plan.

Deeper documentation

1. Architecture
2. AI providers
3. macOS manual test plan

License

MIT