--- title: "Sanity Hosted MCP Server: review, setup & limitations · Undominated.ai" canonical: https://undominated.ai/mcp-servers/sanity-hosted/ description: "Hosted Sanity endpoint where an assistant runs GROQ queries, patches documents, manages releases and explores schema with OAuth or token auth." --- # Sanity Hosted MCP Server: review, setup & limitations · Undominated.ai > Hosted Sanity endpoint where an assistant runs GROQ queries, patches documents, manages releases and explores schema with OAuth or token auth. [← Explore all mcp servers](/mcp-servers/) CONTENT MANAGEMENT / Sanity # Sanity Hosted MCP Server Hosted Sanity endpoint where an assistant runs GROQ queries, patches documents, manages releases and explores schema with OAuth or token auth. [See setup guidance ↓](#setup)[Original source ↗](https://www.sanity.io) SOURCE REVIEW Reviewed 2026-10-13 Evidence 2 linked sources Publisher Sanity Licence Not established [Read what was—and wasn’t—checked ↓](#review) “The Sanity Model Context Protocol (MCP) server enables AI assistants like Claude Code and Cursor to interact directly with your Sanity projects.” [Sanity · upstream description ↗](https://www.sanity.io/docs/ai/mcp-server) Our analysis follows below. 01 / THE REASONING ## Why this made the selection. - Publisher-hosted structured-content surface: schema-aware GROQ, document patching and release management in one endpoint. - OAuth by default with token-based auth as an explicit alternative, and tool availability follows project permissions. ### A good fit for - Querying content across datasets with schema context the assistant did not have to be handed. - Preparing document patches and release operations against a Sanity project. ### Weigh up before choosing - Tools can create, patch, publish and unpublish documents, discard drafts, deploy schemas and Studios, create projects (initialised with a dataset and API tokens) and datasets, change dataset access settings and add or delete CORS origins; scope the project, dataset and token role first. - The publisher archived the local @sanity/mcp-server package in favour of this hosted server; the local package was not selected. - generate_image and transform_image consume Sanity AI credits on top of project entitlements; the docs suggest disabling them in the client to avoid credit use. - Sanity's MCP documentation moved from https://www.sanity.io/docs/mcp, which now redirects to https://www.sanity.io/docs/ai/mcp-server (checked 2026-10-13). 02 / THE REVIEW RECORD ## What we actually inspected. Source review has boundaries. A clear record is more useful than a “safe” badge. ### Material inspected - Sanity MCP documentation at https://www.sanity.io/docs/ai/mcp-server, read 2026-10-13 (HTML and Markdown export): hosted endpoint, CLI quick install, per-client setup, OAuth and token authorization, the Available tools list, AI credit usage and troubleshooting. - Archived local repository README at revision 64b95d18bb94eebdc62d26cd24abf53c04b8e955 (the default-branch head on 2026-10-13; GitHub reports the repository archived): deprecation notice pointing to the hosted server, which it describes as Streamable HTTP with OAuth. ### Our findings - The documented deployment is remote-only against Sanity infrastructure; the archived local package is not the selectable surface. - Write-capable content operations follow the authenticated identity's project, dataset and token-role permissions; with an Authorization header set, the server uses the API token instead of OAuth. - patch_documents saves edits to the draft or release version, never directly to published content; publish_documents then publishes drafts. create_release records releases but cannot schedule, publish, archive or delete them. - run_sanity_cli runs a limited subset of Sanity CLI commands without a shell or filesystem access. - OAuth sessions typically expire after 7 days; the CLI configure command instead writes a generated token for the logged-in CLI user into the editor's configuration. - configExample is the docs' Cursor mcp.json block. Corrected 2026-10-13: the earlier example and upstream description were not verbatim from the docs. ### Not established by this review - Server connection, authentication and tool execution were not tested. The review applies to the material and revision named here. A newer upstream release can change its behavior. 03 / PUT IT TO WORK ## Connect a server deliberately. [Upstream setup instructions ↗](https://www.sanity.io/docs/ai/mcp-server) - Configure the client against https://mcp.sanity.io, preferably with the Sanity CLI setup command (npx sanity@latest mcp configure), which detects Cursor, VS Code and Claude Code. - Authenticate with OAuth or a least-privilege API token scoped to the intended project and dataset. ### Before you start - A Sanity account with access to the target project and dataset. ### Documented configuration example Use the client and authentication context described in the setup guidance. Configuration formats vary between hosts. { "mcpServers": { "Sanity": { "type": "http", "url": "https://mcp.sanity.io" } } } Copy configuration ↗ ### Compatibility MCP clients supporting Streamable HTTP remote servers; the docs give setup for Claude Code, Cursor, VS Code, OpenCode, v0, Lovable and Replit, and mcp-remote for clients without remote support. · A Sanity account with project access, via OAuth, the Sanity CLI's generated token, or an API token. ### Transports Streamable HTTP ### Sanity account / OAuth or API token - Reads schema, datasets, documents, Media Library assets and project resources allowed by the identity. - Enabled tools may create, patch, publish and unpublish documents, manage releases, deploy schemas and Studios, create projects and datasets, and add or delete CORS origins. - OAuth by default; an Authorization header with an API token replaces OAuth and the token's role then bounds every tool call. ### Cost model Sanity project entitlements apply; flagged AI-inference tools additionally consume AI credits. 04 / FOLLOW THE EVIDENCE ## The source trail. Our notes are separate from the original resource. Check upstream before adopting a new version. - [Sanity MCP documentation ↗](https://www.sanity.io/docs/ai/mcp-server) Checked 2026-10-13 https://www.sanity.io/docs/ai/mcp-server Supports: summary, upstreamDescription, install, compatibility, access, transports, configExample, whySelected, bestFor, limitations, review - [Archived local package notice ↗](https://github.com/sanity-io/sanity-mcp-server/blob/64b95d18bb94eebdc62d26cd24abf53c04b8e955/README.md) Checked 2026-10-13 https://github.com/sanity-io/sanity-mcp-server/blob/64b95d18bb94eebdc62d26cd24abf53c04b8e955/README.md Supports: limitations, transports, review KEEP COMPARING ## Other approaches to consider. Related by category or shared topics. These are alternatives to inspect, not a measured quality order. [### Contentful MCP ↗ Read and manage Contentful entries, assets, content types and environments through its local MCP server.](/mcp-servers/contentful/)[### Contentstack MCP ↗ Expose selected Contentstack API groups to a local stdio client with service-specific credentials.](/mcp-servers/contentstack/)[### Sanity hosted MCP ↗ Query schemas and content, stage document changes and publish through Sanity-hosted MCP using an authenticated account.](/mcp-servers/sanity/) [AI Tools ↗](/tools/)[Skills ↗](/skills/)[Agents ↗](/agents/)[MCP Servers ↗](/mcp-servers/)[Workflows ↗](/workflows/) ## Continue your investigation - [Inspect resource evidence](/tools/) - [Choose a host](/agents/) - [Inspect reusable workflows](/skills/) - [Use Undominated data](/api/)