Nexvo MCP Server - Connect Nexvo to Claude

Nexvo speaks the Model Context Protocol, so an AI assistant can build and publish your Shopify product pages directly. Connect once and Claude can import a product from a URL, lay out the page, write the copy, generate the imagery, and push it live — without you leaving the conversation.

There are two ways in. The connector is the one most people want; API keys exist for scripts and self-hosted clients.

Connect with Claude

No API key and no configuration file — just a URL. The Nexvo MCP is available on the Starter plan and above. Add this URL as a custom connector:

In Claude, open Settings and go to Connectors.
Choose Add custom connector and paste the URL above.
Click Connect. Claude sends you to Nexvo to sign in.
Approve the access request on the Nexvo consent screen.
You land back in Claude with the Nexvo tools available.

Nexvo never sees your Claude account, and Claude never sees your Nexvo password. The connector uses OAuth 2.1 with PKCE — Claude receives a scoped access token that you can revoke at any time from your Nexvo account.

Connect with an API key

For scripts, self-hosted MCP clients, or direct REST calls, authenticate with a personal API key instead. Same plan requirement as the connector — Starter and above.

Open the account menu in Nexvo and choose API Keys.
Create a key and copy it — it is shown once and stored only as a hash.
Send it as a bearer token on every request.

Every tool is also a plain REST endpoint under /api/mcp/, running the same code, auth, credit accounting, and rate limits as the connector:

curl https://nexvo.app/api/mcp/credits \
  -H "Authorization: Bearer nx_your_key_here"

Keys are scoped to your account and respect the same row-level security as a normal session. If you downgrade to Free the key stays in place but stops working until you upgrade again. The same applies to the connector — it stays connected, and access resumes the moment you are back on a paid plan.

What the assistant can do

Twenty-five tools, in four groups. Reads and content writes are free; only image generation spends credits.

Read

Free. No credits, no side effects.

list_stores: List your stores, with search and pagination.
get_store: Full store detail — pages, Shopify connection, preview URLs.
get_page_content: Current content of every section on a page.
get_section_schemas: What sections a page has and which fields each one takes, including appearance settings and the format of icon and avatar fields. Call this before set_content.
list_section_templates: Every section template available, and which of them can be added to a given page.
list_media: What an image or avatar field can be filled with: avatar presets, the monogram form, plus the product and generated images already on the page.
get_credits: Credit balance and subscription tier.
get_publish_status: Whether a page finished publishing, plus the live Shopify URL.
get_image_status: Poll an image or creative generation job until it is ready.
list_creatives: Ad creatives for a store, newest first.

Build

Writes to your store. Still free — no credits.

import_product: Pull structured product data from a URL (AliExpress, Shopify, or any product page).
create_store: Create a store and its product page from imported data.
create_page: Add another product to an existing store. Each page keeps its own persona, angle and offer.
set_content: Write copy, item lists, and settings into a page. Stores exactly what the assistant wrote — no AI runs here.
update_store: Brand, palette, fonts, persona, marketing angle, offer, image style.
update_page: Page-level metadata such as the title.
duplicate_page: Clone a page for A/B testing, including the Shopify product.
add_section: Add a section the preset did not include — a comparison table, an ingredients breakdown, a FAQ.
remove_section: Take a section off a page, discarding its content.
reorder_sections: Set the order the sections appear in.
set_theme: Repaint one page or a whole store in a single call.

Generate

Spends credits — 3 per image. Deducted on success only.

generate_image: AI image for one section field, written back into the page.
generate_logo: AI brand logo, saved to the store.
generate_creative: Ad creative for paid social, mixing store context with your prompt.

Publish

Pushes to your live Shopify storefront.

publish: Render the page, upload theme assets, create or update the Shopify product, return the live URL.

A typical run

Ask for a store in plain language and the assistant works through the pipeline:

import_product pulls the title, images, price, and description from the product URL you pasted.
create_store creates the store and its product page from the category preset. Sections start empty.
get_section_schemas reports which sections exist and what copy each field wants.
set_content writes the copy back in. The assistant wrote that copy itself — this tool runs no AI and costs no credits, it only stores what it was given.
generate_image starts each image and returns a prediction ID. Images are asynchronous: poll get_image_status until it reports ready, roughly 5 to 30 seconds. The URL is written into the section automatically.
publish renders everything, uploads the theme assets, and creates or updates the Shopify product. get_publish_status returns the live URL once it lands.

Publishing a home page replaces your Shopify theme homepage. Product pages are additive and safe — home pages are not, so create_store only makes one when you explicitly ask.

Credits and rate limits

Image generation costs 3 credits per image, and applies to generate_image, generate_logo, and generate_creative alike. Credits are deducted on success only — a failed generation costs nothing. Everything else, including all the copywriting, is free. Call get_credits to check the balance before a long run.

Rate limits apply per credential, not per IP:

import_product20 / minThe scraper is expensive to run.

generate_* · publish · duplicate_page120 / minCredits are the real throttle; this only stops runaway scripts.

Everything else120 / minReads and content writes.

All tools combined600 / minAbsolute ceiling across every call.

A throttled call returns 429 with a Retry-After header telling you how many seconds to wait.

When something goes wrong

A tool is missing, or Claude says it cannot do something listed here: Your client caches the tool list from when it connected, so tools added since then will not appear. Disconnect and reconnect the connector. To check, ask your assistant what the Nexvo server reported on connect — it states how many tools this build exposes, and you can compare that with what your client offers.
401 — Missing or invalid Authorization header: No credential reached the server. For a connector, disconnect and reconnect to refresh the token. For an API key, check the header reads "Bearer nx_..." with the prefix included.
401 — Invalid or revoked API key: The key was deleted, or it belongs to a different environment. Keys issued on a development instance do not work against nexvo.app.
403 — tier_required: The Nexvo MCP is available on the Starter plan and above, and this account is on Free. Upgrade at nexvo.app/plans and the same connector or key starts working — there is nothing to reconnect.
429 — Rate limit exceeded: Wait the number of seconds given in retry_after_seconds. Batching many images at once is the usual cause.
An image never finishes: Keep polling get_image_status. It reports failed rather than hanging when a generation genuinely dies, and no credits are charged for it.

Still stuck? Email admin@optimaier.com.