Run Beaconed from your AI assistant.
Beaconed ships as an MCP server and a CLI — both live on npm today. Your assistant finds weak product listings, rewrites them, and applies the changes you approve, without you opening the dashboard.
Read https://beaconed.ai/agent-setup and follow ALL the instructions to install the Beaconed MCP server and configure yourself to optimize my store's product listings. Ask me before making any config changes.
Your agent handles the mechanics. You just confirm.
The whole install — package, account, config, verification — runs in one back-and-forth with your assistant. The only step you handle yourself is creating a Beaconed account and copying your API key.
Your agent installs the package.
It runs npm install -g @beaconed/mcp for you. No editing dotfiles.
It walks you through signup.
Don’t have a Beaconed account yet? Your agent sends you to beaconed.ai/signup and asks you to copy your API key back. The account is the one part it can’t do for you.
It configures your AI client.
Your agent edits its own config to register the Beaconed MCP server, with your API key in place. It asks permission before writing the file.
It verifies the connection.
Your agent calls a read-only Beaconed tool, confirms it’s wired up, and reports back. Then you can just talk to it about your catalog.
Every Beaconed tool, available to your agent.
Not a curated subset. Two dozen tools that cover the whole Beaconed API — read the catalog, queue rewrites, approve, apply, manage webhooks.
See the catalog
Read access to everything Beaconed knows about your store.
- List products — filter by status, score, letter grade, or title
- Get full product detail with images and score history
- Pull readiness scores across the whole catalog
- Read account settings — brand voice, target keywords, auto-push
Generate rewrites
Beaconed's AI rewrites titles, descriptions, and structured data — the work the optimizer was built to do.
- Queue an optimization for one product
- Bulk-optimize many products in a single call
- List every optimization across the catalog or per-product
- Inspect any optimization — original beside the AI rewrite
- Recalculate readiness scores on demand
Review and ship
Nothing reaches your live store without an approve and apply step.
- Approve a pending optimization, or reject with a reason
- Apply approved rewrites — the moment your store actually changes
- Revert an applied optimization if it isn't working out
- Destructive actions are flagged — your agent asks first
Keep systems in sync
Wire Beaconed into the rest of your stack with webhooks.
- Create, update, and delete webhook subscriptions
- Browse the catalog of available event types
- Send a test event to verify the pipe is wired up
Manage the catalog itself
When Shopify isn't the source of truth.
- Create products from external (non-Shopify) data
- Update product fields directly
- Trigger a Shopify sync to pull the latest data into Beaconed
The full tool reference is on the docs site.
OpenAPI spec, input/output schemas, every tool’s destructive flag.
An MCP server and a CLI.
Same tools, same API key, same approve-before-apply rules. Your agent uses whichever fits. Both published on npm and ready to install.
MCP server
@beaconed/mcp
Drop Beaconed into Claude Desktop, Cursor, Claude Code, or any MCP client and operate your catalog in plain English.
Men’s category averages 71. 8 products are below 60 — want me to queue rewrites for those?
-
Adds itself to your client via
claude mcp addor one config line - Read, optimize, review, and apply by chatting — no UI
- Apply, revert, delete are flagged destructive — the agent confirms first
CLI
@beaconed/cli
The same capabilities as plain commands. The right surface when an agent wants to script, pipe, or drop into a CI step.
-
Install:
npm install -g @beaconed/cli -
Human-readable tables, or
--jsonfor piping - Drops into a Dockerfile or CI step the same as any other tool
Hand over the work. Keep the keys.
An agent runs the workflow. You stay in control of what reaches your live store. Both, by default — no extra setup, no “trust mode” toggle.
Nothing reaches your storefront without you.
Optimizations live in Beaconed as drafts until you say apply. Your agent stages the work; you (or an opt-in auto-publish setting) decide what ships. Reverting is a single tool call away.
Risky tools are clearly marked.
Apply, revert, and webhook-delete are flagged as destructive in the MCP manifest, so any reasonable assistant pauses for confirmation before touching them — and you'll see exactly which tool it wants to call.
Local install, your own API key.
The MCP server runs on your machine and authenticates directly against Beaconed with the API key you control. No third party in the path. Revoke the key, the door closes.
Auto-publish is opt-in per account — for merchants who want the agent to ship low-risk rewrites without asking.
Two npm commands and a config snippet.
The paste-a-prompt path is the fast way. But the install really is just an npm package and a config file — here’s the manual recipe.
Install the package
The MCP server and the CLI are separate npm packages — pick the one you want first, or install both. They share the same API key.
npm install -g @beaconed/mcp
npm install -g @beaconed/cli
Register the MCP server
Drop this block into your client’s MCP config. Grab your API key from your API keys page.
{
"mcpServers": {
"beaconed": {
"command": "npx",
"args": ["-y", "@beaconed/mcp"],
"env": {
"BEACONED_API_KEY": "bcn_live_xxxxxxxxxxxx"
}
}
}
}
Paste the prompt. Your agent does the rest.
Paste it into Claude, Cursor, or whichever assistant you use. The setup runs itself — your part is creating a Beaconed account and copying an API key.
Read https://beaconed.ai/agent-setup and follow ALL the instructions to install the Beaconed MCP server and configure yourself to optimize my store's product listings. Ask me before making any config changes.