Add Sprites to GitHub Copilot
Available on Copilot in VS Code, and Copilot CLI. Any Copilot plan that includes agent mode. You need a Sprites account with Shopify already connected — the connector acts on that connection, so GitHub Copilot never handles a Shopify credential.
Open the MCP configuration
Run MCP: Add Server from the Command Palette and choose Workspace or Global. Workspace writes .vscode/mcp.json, which is worth committing so the team shares the same connector; Global keeps it to you across every project.
VS CodeCommand PaletteMCP: Add ServerWorkspaceAdd Sprites as an HTTP server
The top-level key is servers, not mcpServers, and a remote endpoint needs an explicit type of http. This is the shape VS Code and Copilot share.
.vscode/mcp.json{ "servers": { "sprites": { "type": "http", "url": "https://mcp.sprites.ai/mcp" } } }Authenticate
VS Code shows an Auth action above the server entry once the file is saved. Run it, sign in to Sprites in the browser window that opens, and approve the scopes.
Spriteswants access on behalf of Copilot- Read your connected ad accounts and site data
- Draft campaigns, creatives and pages for your approval
- Apply only the changes you approve
AllowCancelSwitch Copilot Chat to agent mode
MCP tools are only offered to the agent. Pick Agent from the mode dropdown in Copilot Chat, then open the tool picker and confirm Sprites is listed.
Agent modeCopilot Chat → mode dropdown
Run the first prompt
Ask GitHub Copilot this. It routes to a Sprites door, which reads your Shopify account and answers — no account to name, no confirmation to give first.
- With Sprites, which products are profitable after ad cost this month?
- Ask Sprites to list every product in the summer collection with price and inventory.
- Use Sprites to create a collection page for "gifts under $50" and draft it in Shopify.
- Have Sprites rewrite the 20 worst-converting product descriptions.
- With Sprites, build Meta creative from our three best-selling SKUs.
Open the Copilot Chat tool picker in agent mode; Sprites appears with its tools, each one toggleable.
What GitHub Copilot can do with Shopify
The same doors, whichever client you are in. Marketing data inside the editor where the tracking code lives, so a conversion event can be checked against what GA4 actually recorded before the pull request is opened.
Profit after ad cost, per product
The join nobody does by hand: store revenue against what Google, Meta and TikTok spent to get it.
Read the whole catalogue
Every product, variant and collection — uncapped, filterable, and grep-ready when you want to work on it.
Publish collection pages
Real pages in the store, with metadata and internal links, arriving as a draft to approve.
Refresh product copy at scale
Descriptions rewritten against what converts, proposed as a set you review before it lands.
Creative from real products
Ads built from the actual catalogue — price, variant, image — rather than a model's guess at your range.
Store blog without a second tool
Posts drafted into the Shopify blog through the same connector as everything else.
How approvals work here
GitHub Copilot renders tool results as text, so a Shopify change comes back written out rather than as a card. Nothing is applied until you approve it, and the approval waits for you in Sprites — open it there and the same run continues.
What it reads
- Products, variants, collections, inventory and pricing
- Orders, revenue and sessions
- Store pages, blog posts and their metadata
- The ad spend attributable to each, from the connected ad accounts
What it can change
- Collection and landing pages, and store blog posts
- Product descriptions and page metadata
- Nothing touches the storefront without an approval
When it does not work in GitHub Copilot
Three things go wrong in this client specifically, and each has a one-line fix.
Sprites is configured but Copilot never uses it.
Chat is not in agent mode. Ask mode does not call MCP tools at all.
The config is ignored.
The top-level key is servers. mcpServers is the shape other clients use and VS Code does not read it.
Teammates do not see the connector.
It was added Globally rather than to the Workspace. Move it to .vscode/mcp.json and commit that file; each person still signs in as themselves.
Shopify and GitHub Copilot: questions
Which URL do I paste into GitHub Copilot?
https://mcp.sprites.ai/mcp. There is one endpoint and every client uses it — every OAuth grant is bound to exactly that address, so a variant of it will not be honoured.
Can GitHub Copilot change my Shopify account without asking?
GitHub Copilot renders tool results as text, so a Shopify change comes back written out rather than as a card. Nothing is applied until you approve it, and the approval waits for you in Sprites — open it there and the same run continues.
Do I need a Shopify API key or developer token?
No. Shopify is connected once inside Sprites through Shopify Admin, and the connector acts on that connection. GitHub Copilot never sees a credential for it.
Why GitHub Copilot for Shopify specifically?
Marketing data inside the editor where the tracking code lives, so a conversion event can be checked against what GA4 actually recorded before the pull request is opened.
Do I need a Shopify Partner account?
No. Connect a custom app you create in your own Shopify admin and paste its access token, or use the standard OAuth connect. Neither needs a Partner account.
Can it change prices or inventory?
No. Pricing and inventory are deliberately out of scope. It reads them, and writes only content: pages, posts, descriptions and metadata.
Keep going
The same GitHub Copilot connector reaches every other channel — nothing else to add.
Shopify, from GitHub Copilot
Paste one URL, sign in once, and the next Shopify question you ask GitHub Copilot gets answered from the live account.