Motozmo MCP Server
# Motozmo MCP Server
Exposes [motozmo.com](https://www.motozmo.com)'s motorcycle repair guides and gear reviews
as MCP tools, so AI assistants (Claude, Cursor, etc.) can search and quote them directly.
**Tools:** `search_guides` · `get_guide` · `list_latest` · `list_topics`
## Auto-updates: yes
Article data is **not** baked into the code. The server fetches the live WordPress
REST API (`/wp-json/wp/v2/posts`) and caches it for 6 hours. When Rehana publishes a
new article, it appears in tool results automatically after the cache refreshes —
no redeploy, no manual re-indexing.
## Run locally (stdio)
```bash
npm install
node server.js
```
Claude Desktop config (`claude_desktop_config.json`):
```json
{
"mcpServers": {
"motozmo": { "command": "node", "args": ["/absolute/path/to/motozmo-mcp/server.js"] }
}
}
```
## Host remotely (Streamable HTTP)
```bash
node server.js --http --port 3000 # serves POST /mcp, GET /health
```
Options, cheapest first:
1. **Render / Railway / Fly.io** — deploy this folder as a Node service
(`npm install && node server.js --http --port $PORT`). Free tiers work for
this workload (tiny traffic, 6h cache).
2. **Own VPS/shared hosting** — any host that runs persistent Node.js 18+.
3. **Cloudflare Workers** — needs a small adapter (workers-mcp); ask before going this route.
Remote servers are what MCP directories list and what other people's assistants can connect to.
## Publish for directories
```bash
npm publish # then directories can reference it via npx -y motozmo-mcp
```
See `SUBMIT.md` for the per-directory submission checklist, and `server.json`
for the official MCP registry (`mcp-publisher publish`).
## Test
```bash
npm test # boots over stdio, lists tools, calls each one against the live site
```
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose: search by keyword, fetch full guide by identifier, list recent guides, and list topic areas. There is little risk of misselection because the verbs and resource targets are distinct.
The set mostly follows a verb_noun pattern (search_guides, get_guide, list_topics), but list_latest deviates slightly by using an adjective instead of a noun. The deviation is minor and still readable.
Four tools is well-scoped for a read-only content browsing server. Each tool maps to a distinct browsing need: search, retrieve, discover latest, and understand topic coverage.
The core read-only lifecycle for content discovery and retrieval is covered. Minor gaps exist, such as no explicit topic-filtered search or pagination controls, but agents can work around these.