One master message, adapted on-brand for every audience — without the voice drifting.
Message Variant Engine encodes your brand voice as a reusable AI skill, then generates persona-, region-, and channel-specific variants from a single master message. Sections you mark as locked (legal language, product names, CTAs, claims) stay verbatim; everything else flexes to the audience. The brand owner defines the system once; anyone on the team produces on-brand variants in seconds.
The design/brand problem it solves: The moment a message has to reach more than one audience, brand voice starts to fragment. A marketer rewrites it for enterprise, a CS lead softens it for churned users, a regional team localizes it for APAC — and three weeks later the brand sounds like five different companies. The rewrite work is invisible, unowned, and impossible to keep consistent by policy alone.
If you only have half a minute, look here:
ai-skills/message-adaptation-skill.md— the AI skill itself: a structured, markdown instruction set that turns brand rules into reliably on-brand output. This is the heart of the project.docs/design-system.md— the brand system as source of truth: voice, tone, type/color/layout and content rules the tool enforces.docs/PRD.md— the product thinking: problem, users, goals, and adoption metrics.docs/architecture.md— how it works: UI surface, the AI/model layer, and the prompt layer that binds them.extension/+ demo — the working Chrome extension (Manifest V3) with a popup UI. (Demo GIF: seedocs/demo.gifonce added.)
Brand & design owners (who define and protect the voice) and non-designers — marketers, customer success, sales, regional teams — who need audience-specific copy without going off-brand.
Message Variant Engine is a Chrome extension (Manifest V3) with a popup UI. The brand owner defines the voice and rules once as a markdown reference framework (see docs/design-system.md and ai-skills/message-adaptation-skill.md). At runtime, the tool composes those rules with the user's input into a structured prompt and sends it to Anthropic Claude (bring-your-own API key). The result is on-brand output that a non-designer can produce without ever seeing the underlying system.
- Input: Master message, target personas, target regions/locales, target channels, and any locked (verbatim) sections.
- Output: One on-brand variant per persona × region × channel combination, with locked sections preserved exactly.
- Clone this repo.
- In Chrome, open
chrome://extensions, enable Developer mode, and click Load unpacked. - Select the
extension/folder. - Open the extension, add your Anthropic API key (stored locally, never sent to any server but Anthropic), and go.
No accounts. No backend. No analytics. See PRIVACY.md.
On-brand output is not luck — it's a designed system:
- The brand voice is encoded, not remembered. Tone, vocabulary, and rules live in
docs/design-system.mdas the single source of truth. - The AI skill enforces it.
ai-skills/message-adaptation-skill.mdis a structured instruction set with explicit constraints, examples, and guardrails — the same discipline you'd apply to a component library, applied to language. - Adaptation is bounded. The tool only flexes what should flex (audience, region, channel); everything the brand must protect stays fixed.
- Anyone can self-serve. Non-designers produce on-brand work because the design system is doing the work for them — invisibly.
This repo is part of a small suite of tools exploring one idea: the best design systems are increasingly invisible — AI skills, brand-encoded tools, and markdown reference frameworks that let anyone produce on-brand work. A concrete example of an AI skill built from scratch, where a markdown-encoded brand voice is the source of truth and non-designers self-serve on-brand output.
Related tools in the suite:
- Message Variant Engine — one message, every audience, on-brand.
- Brief Engine — scattered inputs → structured, on-brand brief.
- Social Copy Multiplier — one message, platform-perfect everywhere.
MIT — see LICENSE.