Summary

The discovery tools are what make the generation usable.

Generating on-brand copy needs IDs the model has no way to guess, so listing style guides, brand voices and audiences — with descriptions attached — is what lets it choose an appropriate tone instead of asking you for a UUID. `apply-style` is the quieter win: text you or the model already wrote gets conformed to the brand rules without regenerating it from scratch.

What it is

A prototype server for the Jasper AI API. What separates it from asking a model directly is that generation is anchored to your account's configuration: a style guide ID, a brand voice tone ID and an audience ID, all fetched from Jasper and passed into the request. It can also take text you already have and conform it to a style guide.

What you get

`generate-content` produces content through Jasper against your style guide, brand voice and audience
`apply-style` takes existing text and rewrites it to a specified style guide's rules
`get-jasper-style-guides` returns the style guide IDs on your account
`get-jasper-brand-voices` returns the brand voices with descriptions and tone IDs, so the model can pick the right one
`get-jasper-audiences` returns audiences with descriptions and audience IDs
Resources at `jasper://styles`, `jasper://brandvoices` and `jasper://audiences`

Requirements

A Jasper API key, set as `JASPER_API_KEY`. The quickest path in Cursor or Windsurf is `npx -y` against the repository; for Claude Desktop there is a `.dxt` extension you drag into the Extensions window and configure with your key. If npx misbehaves under nvm, clone it, `pnpm install`, `pnpm run build`, and point the config at `dist/index.js` with an absolute node path (package `jasper-ai-mcp-prototype`, version 1.0.0). It is labelled a prototype by the authors.