vibe coded with ❤️
Concept app · v0.1.0 · Notarized

Changy

A native macOS app that runs your ITIL change-management process end to end — evaluate a draft change plan against your own policy, fill the gaps through a guided interview, and generate audience-ready communications. All on a local LLM.

Concept app. Changy explores what on-device AI could do for ITIL change management. The core flow runs end to end, but it's a prototype, not a supported product — review everything it produces before it goes anywhere near a real change board. It has been tested in a demo environment only, not in production.
Changy Evaluation screen: 2 of 7 required criteria satisfied, 1 partially satisfied, 4 missing, each with a specific citation from the draft plan

From draft plan to done deal

🔍

Gap Evaluation

Ingests your change policy, draft plan, supporting documents, and even whole codebases. Extracts a custom rubric from your own policy — or falls back to a generic ITIL v4 checklist — and scores the plan with cited, specific findings. Can optionally check claims against live web search.

💬

Guided Interview

Turns every gap the evaluation found into a concrete follow-up question, then merges your answers back into a fully revised plan.

📣

Audience Communications

One click to generate a distinct, tone-appropriate write-up for end users, management, stakeholders, and colleagues from the finished plan — ready to copy into an email or chat tool.

🔌

Enterprise Integrations

Optional read-only lookups against Jira, Confluence, Jamf Pro, and Microsoft Intune during evaluation — e.g. checking related tickets or the actual scope of managed endpoints a change affects. Off by default, independently configured.

🧠

Local or Cloud LLM

A bundled, 4-bit Qwen2.5-7B model runs fully offline via MLX — no first-run download. Or point Changy at any OpenAI-compatible API instead, switchable per deployment.

🔐

MDM Configurable

Every setting — which backend to use, API keys, which integrations are enabled — can be set in-app or centrally forced via a standard macOS Configuration Profile, with the usual "set by your organization" locked UI.

Configuration & Examples

The full Managed Preferences reference — every config key, the local vs. OpenAI-compatible backend, Jira/Confluence/Jamf Pro/Intune setup — plus a worked example using the two sample documents shipped with Changy.

📖 Read the Documentation

Where things stand

System Requirements

macOS 26 (Tahoe) or later.

The bundled on-device model requires Apple Silicon; using an OpenAI-compatible API backend instead has no such requirement.

Build Status

Version 0.1.0. Signed with a Developer ID Application certificate and notarized by Apple (stapled — spctl -a -vv reports accepted, source=Notarized Developer ID).

Distributed as a signed, notarized .zip — download, unzip, and open. No Gatekeeper workaround needed.

What's proven

Evaluation, interview, and communications generation are implemented, tested, and verified end to end with real runs against the bundled local model.

What's not yet validated

The Jira, Confluence, Jamf Pro, and Intune integrations, plus the web search tool, are implemented and unit-tested against mocked responses only — not yet checked against real instances.

Want to try the concept?

Concept app · Free · Signed & Notarized · macOS 26+ · 3.6 GB (bundled model included)

Download Changy