MCP Basics

How Popmelt MCP works with app-first creation and model updates.

tl;dr Popmelt MCP gives AI assistants tools to discover models and patterns, reference exact slices of context, create app artifacts from specs, sample webpages, and update private models.

What is MCP?

The Model Context Protocol lets AI assistants call external tools safely. Popmelt uses MCP so your assistant can fetch model data and generate UI artifacts in-chat.

Available tools

Popmelt provides six MCP tools:

Tool	What it does	Who can use it
connect	Shows account status and active model guidance	Everyone
discover	Finds models, patterns, and artifacts (ID-first)	Everyone
reference	Fetches model/pattern context with optional path queries and opt-in code access	Everyone
create	Creates an app/artifact from a model + `spec` (data-only response)	Everyone
sample	Samples tokens/components from public webpages	Iterate/Scale subscribers
update	Creates/edits private models with `modelDefinition` and `changes[]`	Create tier and above

Typical flow

connect to establish context.
discover models/patterns and capture IDs.
reference the exact data needed (path when possible).
create with a spec to render an app/artifact.
Only request code when needed: reference { targetType: "model", modelId: "...", includeCode: true }.
update private models when you need token/component/page-code changes.

Why IDs + path matter

IDs avoid name collisions and are preferred for client LLM calls.
Path queries keep payloads small (/principles, /components/button, /page/components).
Lean defaults reduce token usage; code is opt-in via reference.

What's next

Better prototypes with less prompting

Start creating