There are various ways to use AI with Siteglide both inside the UI and externally (MCP).

Siteglide MCP

Siteglide has two MCP (Model Context Protocol) servers. MCP is an open standard that lets an AI assistant use tools from other services. Both Siteglide MCPs connect an AI assistant to your Siteglide sites, but they do different jobs.

• The Cloud MCP works with your site content and data, such as CRM users, form cases, products, orders, WebApps, Modules and pages. It's a hosted connector. You add it to your AI app and sign in with your Siteglide account. There's nothing to install.

• The Local MCP works with your code. It runs in your code editor alongside Siteglide CLI and helps you validate Liquid, GraphQL and YAML, test against staging, sync safely and read logs.

Which one do you need?

Frequently asked questions

Which one should I choose?

Choose the Cloud MCP if you want an AI assistant to work with your site data without a local project. For example, it can look up a CRM user and their orders, create or update a product, add a form case for a user, or edit WebApp items and pages.

Choose the Local MCP if you're writing code. In your code editor, the assistant can check Liquid, GraphQL and YAML before it saves files, test Liquid against staging, run GraphQL queries and read recent logs.

Neither MCP provides a chatbot or a widget for your website, and neither includes analytics or reporting tools.

How do they handle sign-in and safety?

Cloud MCP: you sign in with your own Siteglide account, so you don't need an API key. The Cloud MCP acts as you. It can reach every site you can access, and within the tools it offers, it can do whatever your Siteglide user can do on those sites, live sites included. Only connect it with an account whose access you're happy for the AI assistant to use. You can see every tool call it makes in Portal > MCP Audit Log. Read more in Cloud MCP.

Local MCP: it uses the Siteglide CLI credentials already on your machine. The assistant refers to environments by name, such as staging, and should never read or paste your tokens. The Local MCP labels each environment as staging or production, blocks Liquid evaluation on production and asks for your approval before a GraphQL change on production. Sync, pull and deploy stay in your hands. Read more in Local MCP.

Can I use both?

Yes. They don't depend on each other. You can connect the Cloud MCP to work with site data and also run the Local MCP in your code editor for code work. Each one uses its own sign-in, and each one only reaches what those credentials allow.