Storesynk is designed to be built by AI. The tags are declarative, so an agent composes a storefront the way it writes HTML. You describe the store; the agent writes tags.
The knowledge
Section titled “The knowledge”The MCP server serves your AI tool the complete tag whitelist, page blueprints, structure rules, and a symptom catalog, kept in sync with the engine. There is no plugin to install. It is the same source as this site’s tag reference, so if a tag is not in the whitelist, the agent knows it does not exist.
The context
Section titled “The context”Through the MCP server the agent sees your actual shop:
- your real products and collections
- your configured bundles, discounts, customer groups, and page gates
- a credentials smoke test
- a validator for every generated page
The agent can also create bundles, mix and match offers, volume discounts, customer pricing rules, and page gates from the conversation (these are Pro features of the app). Everything is created inactive, and the agent asks first.
What a session looks like
Section titled “What a session looks like”Connecting the server puts the tools within reach; naming Storesynk in the prompt makes the agent pick them up:
“Build me a store using Storesynk: homepage with my top collections, product pages with variant pickers, a filterable collection page, and a cart drawer. Astro, statically rendered.”
The agent pulls the store briefing, writes the pages against your real product handles, and validates the markup. Follow-ups like “add a volume discount widget” work the same way.
Connect your tool
Section titled “Connect your tool”- Claude Code: one command in the terminal
- Claude Desktop: custom connector
- Cursor: one-click deeplink or
mcp.json - Lovable: connectors panel with OAuth
Then see what the AI can do.
Prefer to write it yourself?
Section titled “Prefer to write it yourself?”The getting started sequence and the framework guides are the manual path, using the same tags.