Personal AI Infrastructure (PAI)

Why it's ranked here

The project shows unusually serious installation engineering for a personal AI setup. Dry runs, conflict scans, backups, atomic configuration writes, and explicit permission gates reduce common setup risks. The verdict is still mixed for casual users because it depends on another capable AI harness, requires Bun, and treats several valuable capabilities as optional components or future work.

What's good

Existing user files win during setup, while settings and hooks merge additively. Installers refuse known development trees, report missing payloads as blockers, and detect case-insensitive skill collisions before mixing directories. Security hooks scan outgoing tool data for credential patterns and frame attacker-writable web, mail, drive, calendar, inbox, and Dropbox content as untrusted data.

Tradeoffs

This is infrastructure around an AI harness, not a standalone assistant or model runtime. Claude Code is the most-tested path despite broader harness detection. Local-model support, granular model routing, remote access, outbound calling, and external notifications remain roadmap items. Some background components rely on launchd, and the documented one-line terminal installer is limited to macOS and Linux.

How to use it well

It best suits people already living in a capable coding harness who want durable personal context, repeatable skills, safety hooks, and automated maintenance across many sessions. Start with core installation and version-controlled backups, then opt into services only when their operational value is clear. It does not replace the underlying harness, provide current local-model execution, or serve as a simple prompt collection.

Observed

License

MIT License

Primary language

TypeScript, with Bash also used by the project

Runtime requirement

Requires Bun and a capable AI coding harness

Packaging

Installs as one self-contained skill bundling the broader library

Install surface

AI-guided prompt installation, plus a one-line terminal shortcut for Claude Code on macOS and Linux

Harness detection

Detects Claude Code, OpenCode, Hermes, Cursor, and OpenClaw

Tool interface

Bun CLI entry points use dry-run and apply modes and emit JSON reports

Platform detection

Environment detection recognizes macOS, Linux, and Windows