--- title: "Sign Customiser MCP Server | Connect Claude and ChatGPT" description: "Connect Claude, ChatGPT, Cursor or Codex to your Sign Customiser store. 39 tools for customisers, pricing, products, orders and analytics, behind permissions you approve and can revoke." url: "https://www.signcustomiser.com/mcp/" lastModified: "2026-08-25" --- # Your sign customiser, inside the AI assistants you already use. Connect Claude, ChatGPT, Cursor or Codex to your Sign Customiser store with one URL. Ask for a new colour, a price change, last month's orders. It does the work in your admin, with permissions you approved and can take back. [Connect an assistant](/help/mcp/guides/connect/)[Add Sign Customiser to your store](/install/) https://web.signcustomiser.com/mcp ## Ask in a sentence. Watch it land in your admin. The assistant calls a named tool with real arguments and gets a real record back. Nothing is scraped, guessed or typed into a form on your behalf. - **39 tools** covering customisers, options, pricing, products, orders, quotes, webhooks and analytics - **44 tools** once you approve the advanced catalogue, which reaches the rest of the platform API - Every tool needs the same permission its /api/v3 counterpart needs [See every tool](/help/mcp/) TOOLcreate\_customiser\_option Live Merchant assistant isError: false ## What an MCP server actually is The Model Context Protocol is the standard way an AI assistant connects to software it does not own. Sign Customiser runs one for your store. Point an assistant at it and your customisers, pricing, products and orders become things it can read and change directly, rather than things you paste into a chat window. It is the same platform your storefront runs on, reached a different way. Nothing new to install, nothing to keep in sync. ## What you can ask it to do Every one of these is a named tool with a documented shape, not a prompt trick. ### Set up customisers Start a customiser from a sign prototype, rename it, change its labels, and take it live or off the storefront when you are ready. list\_customiser\_prototypes create\_customiser set\_customiser\_status ### Edit every option family Colours, fonts, sizes, materials, backboards, backlights and mountings. Add records, reorder them, change them, remove them. create\_customiser\_option reorder\_customiser\_options ### Change pricing Read the pricing document a customiser sells at, then replace or merge it. Prices are integer minor units with an explicit currency, so nothing rounds behind your back. get\_customiser\_pricing update\_customiser\_pricing ### Create products and orders Record a product built from a customiser, or an order against it, and hand it off for production the same way your storefront does. create\_integration\_product create\_order ### Look things up Orders, quotes and design enquiries, your subscription, and an order and revenue summary over 7, 30, 90 or 365 days. list\_orders list\_quotes get\_analytics\_summary ### Wire up the rest of your shop Manage the webhook subscriptions that tell your other systems about new orders and products, and search the Sign Customiser documentation without leaving the chat. create\_webhook\_subscription search\_docs ## Connecting takes one URL No API key is created, pasted or stored anywhere in this flow. 1. 01 ### Paste one URL Your assistant asks for a server address. Give it https://web.signcustomiser.com/mcp. There is no API key to create, copy or store. 2. 02 ### Choose a store and approve permissions Sign Customiser opens a consent screen in your browser. Pick the store the assistant acts for, then approve permissions written in plain language, such as View customisers and Manage pricing. 3. 03 ### Watch what it does Tools & Settings > MCP server in your admin lists every connected assistant, how many calls each tool made, and every change a connection has written. 4. 04 ### Revoke in one click Revoking takes effect immediately and does not need the assistant's cooperation. [Setup steps for Claude, ChatGPT, Cursor and Codex](/help/mcp/guides/connect/) ## You decide what it is allowed to touch An assistant with access to your storefront pricing is worth being careful about. The permissions are the product, not a footnote. ### Permissions you chose, checked on every call A connection only holds the permissions you ticked. Ask an assistant to do something you did not approve and the tool answers with the name of the permission it needs, rather than doing it anyway. ### One store per connection A connection is bound to one store for its whole life, and the store comes from the token rather than from anything the assistant sends. Connecting a second store means connecting again and picking it. ### Dry runs before writes Every write that supports it takes dry\_run: true and returns the same verdict without changing anything. Deletions, order creation and product creation also need an explicit confirmation. ### A record you can read Every change is logged with the tool that made it, the assistant that called it, the member who approved the connection, and whether it succeeded. ## Building the client rather than using one? The endpoint, the OAuth discovery documents, the scope list and a copy-paste config for every client we have tested are all on one page. - [Connection referenceEndpoint, discovery URLs, scopes, client configs, first call](/mcp/connect/) - [Tool referenceA page per tool, with arguments, results, errors and examples](/help/mcp/) - [Authentication guideDiscovery, dynamic registration, PKCE, token lifetimes, revocation](/help/mcp/guides/authentication/) - [mcp-tools.jsonThe whole surface as one machine-readable manifest](/help/mcp-tools.json)