Manage Your Catalog by Chat: The Describely Agent and MCP Server

Agent vs. MCP: which one do I want?

Describely gives you two ways to work through conversation instead of clicking around.

The Agent lives inside Describely and can run almost everything in the app for you.
The MCP server connects Describely to outside AI tools like Claude and Cursor, so you can run core catalog actions without leaving the assistant you already use. This article covers what each one does and when to reach for it.

Both let you talk to Describely in plain language. The difference is where you're working and how much you can do.

	Describely Agent	Describely MCP Server
Where it runs	Inside the Describely app	Inside Claude, Cursor, Windsurf, or VS Code
Best for	Doing anything in your catalog hands-free	Core catalog actions from the AI tool you already use
Scope	Full app (generate, enrich, optimize, publish, images, rulesets, and more)	A focused set of actions: find, create, generate, enrich, bulk jobs
Setup	Nothing to install, it's already there	One-time install with your API key

Short version: use the Agent for the full workflow inside Describely. Use MCP when you'd rather drive Describely from Claude or another AI tool.

The Describely Agent

The Agent is an AI assistant built into Describely. Ask it for what you want in plain language and it handles the steps for you, the same way you'd brief a teammate.

You'll find it on the Ask Agent button inside the app. You can click its suggested actions or just type to kick off tasks.

What the Agent can do

Find and view your catalog

Look up products and catalogs, and pull up full product details on request.

Add products

Create a single product from pasted text, a scraped URL, or the product form.
Import products in bulk from CSV or Excel, Shopify, Google Sheets, or your other connectors.

Generate content

Write product content inline in chat for one product, or kick off a bulk job for many.

Fill data gaps

Enrich products by web-searching specs, dimensions, materials, and images, for a single product or in bulk.

Optimize for search

Improve products for AI and traditional search (GEO and SEO), single or bulk. Requires Search Optimization

Get content live

Publish to a connected store, export to CSV or Excel, or re-sync a catalog.

Manage images

Resize, remove backgrounds, and create AI image variations.

Manage rulesets

Create a ruleset from scratch, a template, a clone, a URL, or a document. Edit it, set a default, or delete it.
Manage brand tones and restricted words.
Experiment with a ruleset by testing it against sample content and iterating before you save.

Stay on top of jobs and settings

View and manage your bulk jobs.
Connect Google Search Console and view your search metrics.
Set up the ShopAssist storefront chatbot widget.
Manage your credits and plan, and get routed to support when you need a human.

Examples to ask the Agent:

"Create a product from this URL and write a description."
"Import the products in this Google Sheet."
"Enrich every product that's missing dimensions, then generate descriptions."
"Run a bulk job to optimize my winter collection for search."
"Publish the approved products to my Shopify store."

Tip: The Agent works best when you tell it the outcome you want, not the clicks. "Fill in missing materials for my whole catalog" beats walking it through each step.

The Describely MCP Server

MCP (Model Context Protocol) is an open standard that lets AI assistants connect to outside tools. Describely's MCP server uses it to give Claude, Cursor, and similar tools secure access to your Describely account. So you can manage your catalog from the assistant you already work in, using your Describely API key.

It covers the core content actions, not the full app. For everything else, use the Agent inside Describely.

What you can do over MCP

Find products: look up a product by name or SKU, and pull full product details with images.
Create a product: build a new product from free-form text.
Generate content: see the available fields, generate AI content for a product, and save it back to the product's fields.
Browse your setup: list your catalogs, list and view your content rulesets, and list your brand tones.
Run bulk jobs: preview how many products a job will affect, then run bulk content generation or bulk data enrichment.
Check progress: list your bulk jobs and their status.

Note: Rulesets and brand tones are read-only over MCP. You can apply and view them, but create or edit them in Describely (or ask the Agent).

Set up MCP

You'll need a Describely API key first.

In Describely, click Account > Business Settings > Security & Access.
Under API Keys, click Create API Key. Your key starts with ds-.
Copy it right away. It's only shown once.
Then run the interactive setup from your terminal: npx @copysmith.ai/describely-mcp-server init

This detects your installed AI tools (Claude Desktop, Claude Code, Cursor, and others), asks for your API key, and configures the tools you choose. The server works with Claude Desktop, Claude Code, Cursor, Windsurf, and VS Code.

If you'd rather configure a tool by hand, the npm package page has the config for each one.

Reminder: Your API key is like a password to your business data. Store it in a password manager or environment variable, never in shared code.

Example things to ask in your AI tool

"List my Describely catalogs."
"Find the product with SKU ABC-123."
"Generate descriptions for my winter collection."
"What rulesets do I have configured?"
"Run bulk enrichment on products missing descriptions."

Note: MCP runs through Describely's API using your API key. For how API usage is measured and billed, see About Describely's API Logic and Tokens.

Which should I use?

Working inside Describely, or need the full workflow (publishing, images, search optimization, ruleset setup)? Use the Agent.
Living in Claude or Cursor and want to run core catalog actions without switching tabs? Use MCP.
You can use both. They share the same catalog and rulesets.

Need help? Email support@describely.ai. We're happy to help you get set up.