Sanity MCP Server – Content and Schema Tools

Overview

Sanity's official MCP server connects compatible AI assistants directly to
Sanity projects. It provides schema-aware tools for content and development
workflows, allowing agents to operate with live project context instead of
relying only on copied documents, local schema files, or static model knowledge.

What the MCP server enables

The hosted server exposes editorial and development-focused tools. Depending on
the authenticated user's role or API-token permissions, an AI agent can:

• Inspect deployed schemas and retrieve detailed field and reference metadata.
• Run GROQ queries against authorized datasets.
• Create, retrieve, patch, and manage supported documents.
• Work with releases and migration-related workflows.
• Assist with schema changes, localization, and content-model migrations.
• Manage selected project resources such as datasets and API keys where allowed.
• Retrieve current Sanity guidance and agent rules for supported workflows.
• Use pagination and token-aware responses for large result sets.

When to use it

Use Sanity MCP when an AI workflow needs current project context or must perform
an authorized content or schema operation. Typical examples include querying
articles, inspecting references, patching content, listing releases, adding
localization, migrating documents to a new schema shape, creating datasets,
and helping an agent build or maintain a Sanity-backed application.

Connection and authentication

Sanity hosts the managed server at https://mcp.sanity.io. It supports remote
HTTP connections and works with MCP-compatible clients. OAuth is the default
authentication method and links changes to the signed-in Sanity user.

API token authentication is also supported by sending
Authorization: Bearer <token>. Token-authenticated tool calls are limited by
the token's role and permissions. Clients that cannot connect directly to a
remote MCP server can use the documented mcp-remote stdio bridge.

Key considerations

Access follows the authenticated user's role or the configured API token's
permissions. Use OAuth for user-attributed changes and scoped API tokens when a
narrower permission boundary is required. Review document mutations, release
actions, dataset changes, API-key operations, and migrations before approval.
Keep tokens outside source control, rotate unused credentials, and confirm that
deployed schemas are current before relying on schema-aware tools. Sanity also
offers a separate read-only Sanity Context MCP server for production agents
that only need filtered dataset access.