Sanity MCP Server: Give AI Agents Access to Your Structured Content Lake

Why Give AI Access to Sanity?

Sanity is the structured content platform with a real-time content lake and customizable Studio. When building features, you switch to Sanity Studio to check document schemas, run GROQ queries, verify dataset states, and debug content references. Each context switch costs focus during frontend development.

An MCP server for Sanity lets your AI agent query documents, inspect schemas, run GROQ queries, and check dataset health directly from Claude, Cursor, or Windsurf through DataFaucet.

What Becomes Possible

Once your AI agent has Sanity access:

• "What fields does the article document type have?"
• "Run a GROQ query to find all published posts tagged with 'engineering'"
• "What references point to this author document?"
• "Show me draft documents that were edited in the last 24 hours"
• "What's the schema definition for the page builder array?"
• "How many documents are in the production dataset?"

How It Works

1. Go to DataFaucet and enter your Sanity Studio URL or manage.sanity.io
2. Authenticate with your Sanity credentials
3. Browse documents, schemas, datasets, and Vision tool queries as you normally would
4. Each API call becomes an MCP tool your AI can invoke
5. Deploy and connect to Claude, Cursor, or Windsurf

The AI agent accesses whatever projects and datasets your Sanity permissions allow.

Security Best Practices

• Viewer tokens — create Sanity API tokens with viewer role, not editor or admin
• Dataset isolation — limit AI access to development datasets, not production
• No write access — prevent AI from creating, updating, or deleting documents
• No deploy permissions — block AI from deploying Studio or GraphQL APIs
• No CORS changes — exclude project settings and security configuration
• Token scoping — use dataset-specific tokens when possible

Use Cases by Role

Frontend developers: Schema context. "What fields does the hero portable text block have? What custom types are defined?" while building components that render structured content.

Content engineers: GROQ debugging. "Run this GROQ query and show results. Why does my filter return empty? What documents match *[_type == 'post' && defined(slug)]?"

Full-stack developers: Reference tracking. "What documents reference this author? If I delete this category, what breaks?" while refactoring content models.

DevOps engineers: Dataset health. "How many documents are in production vs staging? What's the last mutation timestamp? Are there orphaned assets?"

Sanity Studio vs AI Agent

Modern Stack Integration

Sanity pairs with Next.js, Remix, Astro, and Nuxt. With DataFaucet MCP servers for each:

• Verify Sanity schema types match your TypeScript codegen output
• Check document state before triggering Next.js ISR revalidation
• Cross-reference Sanity GROQ query results with frontend rendering bugs

Getting Started

Set up read access for your AI agent:

1. Create a Sanity API token with viewer permissions:

- Go to manage.sanity.io → project → API → Tokens

- Create token with Viewer role

- No editor, deploy, or admin permissions

1. Point DataFaucet at Sanity Studio or manage.sanity.io
2. Browse documents, run GROQ queries, and inspect schemas as normal
3. Deploy and connect to your AI coding tool

Related: Contentful MCP Server for Contentful, Notion MCP Server for Notion, Best MCP Servers for the full directory.