rivlwatch-mcp
v1.1.0
Published
MCP server for RivlWatch — track Shopify rivals' sales, read their Meta ads libraries (creatives, copy, reach, media), browse your saved-ads swipe file, and discover winning creatives from any MCP client (Claude Desktop, Claude Code, OpenClaw, etc.).
Maintainers
Readme
RivlWatch MCP server
Bring your RivlWatch competitor intelligence into any MCP client (Claude Desktop, Claude Code, OpenClaw, etc.). Ask in plain language — "which of my rivals sold the most today?", "what is allbirds running on Meta right now?", "summarize the hooks in my Q4 Winners collection" — and it answers from your live RivlWatch account.
Full public docs: https://rivlwatch.com/mcp
Tools
| Tool | What it does | Cost |
|------|--------------|------|
| whoami | Confirm the connection + show the signed-in account/plan | free |
| list_my_stores | Your tracked stores with today's revenue, orders, products | free |
| store_creatives | One tracked store's full Meta ads library: copy, days running, reach, media | free |
| saved_ads | Your saved-ads swipe file + Collections + labels + cached AI reads | free |
| saved_tiktok_finds | Your saved TikTok Shop products + monitored TikTok videos | free |
| my_top_ads | Reach-ranked top creatives across your tracked stores (or the global pool) | free |
| ad_audience | One ad's EU demographics: gender split, age bands, top countries | free |
| winning_creatives | The global winning-creatives discovery feed (filter by window/format/country/AI) | free |
| track_store | Start tracking a store by URL (Shopify, Cloudflare-Shopify, or ads-only) | uses a plan store slot |
All calls inherit your subscription server-side — the MCP can never do more than
your account can, a lapsed plan returns a clear error, and all key traffic
shares one per-account rate budget (HTTP 429 + Retry-After when exceeded).
track_store is the only tool with a side effect; everything else is read-only.
Setup
Get an API key: RivlWatch → Settings → API keys → Create API key. Copy the
rivl_…token (shown once).Install deps (once):
cd mcp-server npm installRegister it with your MCP client.
Claude Desktop — edit
claude_desktop_config.json(Settings → Developer → Edit Config):{ "mcpServers": { "rivlwatch": { "command": "node", "args": ["/ABSOLUTE/PATH/TO/mcp-server/index.js"], "env": { "RIVLWATCH_API_KEY": "rivl_your_key_here" } } } }Restart Claude Desktop. You'll see the
rivlwatchtools available.Claude Code —
claude mcp add rivlwatch -e RIVLWATCH_API_KEY=rivl_… -- node /ABS/PATH/mcp-server/index.js
Verify
RIVLWATCH_API_KEY=rivl_… npm run smokeLists the tools and calls each read tool against the live API. Point at a local
dev server with RIVLWATCH_API_BASE=http://localhost:3007.
Media URLs
Tool responses link ad creatives two ways: archived media under
https://rivlwatch.com/api/media/... is durable — the endpoint requires your
API key (Authorization: Bearer rivl_…) or a logged-in browser, and resolves
to an unguessable storage URL; live fbcdn.net URLs are public but expire
after ~4 days.
Config
| Env | Default | Notes |
|-----|---------|-------|
| RIVLWATCH_API_KEY | — | required — your rivl_… key |
| RIVLWATCH_API_BASE | https://rivlwatch.com | override for local/staging |
Pairs with the RivlWatch skill
Drop rivlwatch-skill/ into your agent's skills folder (e.g.
~/.claude/skills/rivlwatch/ for Claude Code, or your OpenClaw skills dir) so
the model knows the best workflows for these tools — see
rivlwatch-skill/SKILL.md.
