Sanity MCP. Query, edit, and manage your entire Content Lake via chat.

You connect your AI agent directly to your Content Lake using this server, and you're gonna have total control over your content backend. Stop copy-pasting IDs or wrestling with APIs in a bunch of separate terminal windows. This whole setup exposes every core function of the Sanity Content Lake as tools, letting your agent talk right to your data.

Querying Your Data Set

Need specific information? You run complex Graph-Relational Object Queries (GROQ) using run_groq_query. This lets you filter and fetch content across multiple records based on criteria like date ranges or required field values. If a simple lookup won't cut it, this tool gives your agent the power to pull exactly what you need from deep within your CMS.

Reading Documents

When you know a document's ID, you retrieve all of its stored data using get_document_details. This fetches every field value for a single record. If you just wanna list everything available in the system, you can use list_unique_schema_types to get every unique data schema type defined. You can also run list_typed_documents to see all existing document types—like 'post' or 'author'—that your CMS knows about.

Managing Documents and Assets

You build content with the create_cms_document tool, which builds and inserts a brand new document into your Content Management System. To change an existing record without messing up everything else, you use patch_cms_document, letting your agent update only specific fields on a document. For deletions, you permanently remove a file using wipe_cms_document; remember, that action can't be undone.

Beyond the text, you manage all your files with list_media_assets. This tool pulls an inventory of every single image and file asset uploaded to your Sanity project. If you need to know how big your data set is for a specific type, use count_entity_nodes to count exactly how many documents exist for that schema.

Raw Power Moves and Schema Auditing

Sometimes the high-level tools aren't enough. For complex, multi-step data transactions or bulk updates across multiple records, you run raw mutations using run_raw_mutation. This tool handles highly technical, low-level data changes. If you want a definitive list of every single schema type present in your dataset—for auditing purposes—you use list_unique_schema_types.

This setup turns your agent into a full-stack CMS power user. You're not just reading content; you're building, managing, and restructuring the entire data lake without ever opening the Sanity UI.