Agent! (AgentiLoop) — native Swift macOS agent that automates dev and UI tasks with 23 LLMs + Apple Intelligence

What the project is (reporting): Agent! is a 100% native Swift 6.4 / SwiftUI macOS application that the README describes as an “autonomous task loop.” It integrates what the project lists as 23 third‑party LLM providers (OpenAI, Claude, Gemini, Mistral, Ollama local/cloud, vLLM, LM Studio, and others) plus on‑device Apple Intelligence. The app claims to be able to read and modify a codebase, build Xcode projects, commit diffs, invoke shell commands (including as root), drive other Mac apps through the Accessibility API, send iMessages, and respond to a spoken wake word. The project also publishes cross‑platform CLIs (Mac/Windows/Linux) and a v0.0.5 release that adds tools such as glob/grep search, web_fetch, todos, AGENTS.md, /undo, custom slash commands, and –json. Evidence of readiness (from the repo) Practical signs the project is usable today: the repository ships precompiled macOS binaries (dmg/zip) that the README says are built by CI and “signed, notarized, and stapled by Apple” with an AgentiLoop Team ID. The README explicitly recommends using the official release binary for a complete experience because ad‑hoc builds (Option B) are ad‑hoc signed and the Launch Agent/Daemon helpers won’t register without a Team ID. There are two documented build paths: an Xcode flow that requires an Apple Developer account (for helper registration) and a command‑line build that runs ad‑hoc but loses helper registration. The README also claims no Electron/NPM, no subscription, and no telemetry, and that every Swift package dependency was authored by the same project author. What