The free half is a retrieval surface — a fixed body of principles and examples the model can pull by slug instead of improvising design advice. The paid half turns that around and scores your work against the same principles. Worth it only if you want that particular doctrine as the standard; it is opinionated by design.
A public MCP endpoint carrying AI Design Blueprint's doctrine: clusters, principles, worked examples and guides, all retrievable by slug or by search. The same repository ships copy-first rule files for Cursor, Windsurf, GitHub Copilot and Gemini CLI for clients that read files rather than call tools.
- clusters.list and clusters.get — the top-level doctrine map, and one category in full
- principles.list, principles.get and principles.search — a principle by slug when you know it, by query when you do not
- examples.get and examples.search, filterable by principle, difficulty and library
- guides.list, guides.get, guides.search and assets.list for the longer-form material
- signals.report and signals.feedback, which write only the structured fields you pass — never prompts, code or file contents
- me.learning_path, me.coaching_context and me.add_evidence once you are signed in
- architect.validate, design.validate and spec.validate on paid plans — code, rendered frontend and written spec each scored against their own set of laws
- team.summarize for usage across all three validator lenses
Nothing for the read tools: point a Streamable HTTP client at https://aidesignblueprint.com/mcp and the retrieval and signal tools work anonymously. The member tools need a signed-in account; the three validators and the team summary need a paid plan. Each validator call can be marked as a private session so the run is not stored.
One command — claude mcp add --transport http aidesignblueprint --scope user https://aidesignblueprint.com/mcp