@merchinary/docs-mcp
An MCP server that exposes Merchinary's product
knowledge — what the Shopify apps Bulk Product Editor and Image Editor can do, and
how to do it. Point any MCP-capable AI client at it to answer "how do I…" and feature-discovery
questions, grounded in the public docs at https://docs.merchinary.com.
The knowledge is the curated documentation itself (54 pages), baked into a static index at build
time — no database, no auth (the docs are public).
Tools
search_docs | Fuzzy/keyword search across the docs. { query, product?, limit? } → ranked pages with snippet + URL. |
read_doc | Full plain-text of one page. { id } (from a search hit) or { url }. |
list_capabilities | Structured map of products → categories → pages. { product? }. Use for feature discovery. |
Products: bulk-product-editor, image-editor.
Use it (no install — remote)
Connect your client to the hosted Streamable HTTP endpoint:
https://mcp.merchinary.com/mcp
Claude Desktop / clients that support remote MCP servers:
{
"mcpServers": {
"merchinary-docs": {
"type": "streamable-http",
"url": "https://mcp.merchinary.com/mcp"
}
}
}
Use it (local — stdio via npx)
{
"mcpServers": {
"merchinary-docs": {
"command": "npx",
"args": ["-y", "@merchinary/docs-mcp"]
}
}
}
Develop
npm install
npm run build:knowledge
npm run build
npm test
npm run dev:stdio
npm run dev:http
npm run inspector
The index is generated from the curated docs in
web-docs/sites/docs-site/docs/ (single source, shared with the in-app Help search). Rebuild
(npm run build:knowledge) and redeploy whenever the docs change.
Deploy (HTTP transport)
Dockerfile builds a stateless image with the index baked in — suitable for Cloud Run, Firebase,
or any container host. Listens on $PORT (default 8080), serving POST /mcp and GET /healthz.
Registry
server.json is the MCP registry manifest,
listing both the npm (stdio) package and the remote (streamable-http) endpoint.