How is this different from connecting Sanity MCP directly?

Connected directly, an agent sees every tool the server exposes and holds the credential itself. Through TrustGate you choose which tools each agent gets, and every call is logged against the agent that made it.

What if the Sanity MCP API changes?

We track the Sanity MCP schema, so the tool catalogue stays current. Tools you have already allowed keep working, and anything new stays denied until you enable it.

Can I give different agents different Sanity tools?

Yes. Access is set per agent: one can get a handful of read tools, another the full set. You grant the whole MCP or single tools within it.

Where can I see what my agents did in Sanity?

Every call is logged against the agent that made it, with the tool it used. Activity can be sent on to your own monitoring.

Do I need to run the Sanity MCP server myself?

No. TrustGate connects to the Sanity MCP server and sits in front of it. Your agent only talks to your TrustGate URL, never to Sanity directly.

Connect Sanity MCP to your AI agents, securely.

Q: Is Sanity MCP secure?

The server itself exposes 54 tools, and some of them can write to your Sanity account. Behind TrustGate an agent only sees the tools you allow, requests and responses are inspected, and every call is logged. Anything outside the allowlist is denied before it reaches Sanity.

One secured endpoint for every Sanity call your AI agents make, scoped per user, inspected, and logged.

Get started for free

Get a demo

Try Sanity MCP now

What your agents can ask once Sanity MCP is connected.

“Run a GROQ query for all articles written by Mark”
“List all releases in this dataset”
“Add localization to my article document type”

“Help me migrate existing content to a new schema shape”
“Which documents reference this Media Library asset?”
“Show me the deployed schema for this workspace”

Select the tools your AI agents can call

54 tools from the Sanity MCP schema.

Showing 54 tools

query_documentsreadQueries documents using GROQ. Takes the GROQ string and optional params. Results are not truncated, so field projections and slices matter.
get_documentreadFetches a single document by its exact ID. A direct lookup only, with no search or filtering.
list_releasesreadLists releases in a dataset, returning active and scheduled ones by default, with a state filter for published or archived.
create_documentswriteCreates one or more draft documents from structured content. Creates drafts unless a releaseId is given.
create_versionwriteCreates a version document for a specific release. Versions are separate from drafts and published documents and are used for scheduled release workflows.
patch_documentswriteUpdates existing documents with client patch operations, applied as one transaction per document, at most 25 documents per call. Edits are saved to the draft or release version, never directly to published content.
create_releasewriteCreates a release for grouping content changes. Does not schedule it: release type and intended publish time are recorded as metadata only.
publish_documentshigh riskPublishes one or more drafts, making them live.
unpublish_documentshigh riskUnpublishes one or more published documents, moving them back to drafts.
discard_draftshigh riskDiscards one or more draft documents, deleting the drafts while leaving published documents intact.
version_discardhigh riskDiscards one or more document versions from a release.
documents_validatereadValidate documents against the deployed schema without modifying them. This tool is in beta. Results may be incorrect or incomplete; review them before making changes.
media_library_listreadLists the Media Libraries in an organization, returning the library IDs the other media_library tools need.
media_library_get_schemareadGets a Media Library's document types, or the full field definitions for one type. The library schema is fixed, with the organization's own asset aspects merged on.
media_library_query_documentsreadQueries a Media Library using GROQ. Assets, collections, folders and aspect definitions are all queryable.
media_library_get_documentreadFetches a single Media Library document by its exact ID.
media_library_list_asset_referencesreadLists the documents across project datasets that reference a Media Library asset. Intended to be called before replacing an asset.
media_library_create_documentswriteCreates documents in a Media Library. Only asset collections can be created; assets themselves come from uploading a binary. Created directly, not as drafts.
media_library_patch_documentswriteEdits Media Library documents within a safe subset: an asset's title and aspects, its CDN access policy, and collection metadata and membership. Title and aspect edits go to a draft; the CDN access policy is published-only and takes effect immediately.
media_library_publish_documentswritePublishes asset drafts to make unpublished metadata edits live, validating the draft against current aspect definitions first.
media_library_discard_draftswriteDiscards asset drafts, throwing away unpublished metadata edits and leaving the published asset intact.
media_library_link_assetswriteMakes Media Library assets usable in project documents, creating a local deliverable copy for images and files and returning ready-to-use field values.
assets_upload_statusreadChecks one to 20 asset uploads by operation ID. Results include asset IDs, URLs and references, and stay available for up to 24 hours.
media_assets_upload_from_urlreadDeprecated. Does not start an upload; use media_library_upload_from_url with the same arguments.
dataset_assets_upload_from_urlwriteStarts an image or file upload to a dataset from a public HTTPS URL, returning an operation ID immediately.
dataset_assets_upload_from_filewritePrepares a one-use URL for uploading an image or file up to 50 MiB to a dataset.
dataset_assets_upload_from_web_chatgptwriteUploads a ChatGPT web attachment or generated file to a dataset, using the temporary download URL ChatGPT web supplies.
media_library_upload_from_urlwriteStarts an image, video or file upload to a Media Library from a public HTTPS URL, optionally adding a version to an existing asset.
media_library_upload_from_filewritePrepares a one-use URL for uploading an image, video or file up to 50 MiB to a Media Library.
whoamireadReturns the currently authenticated Sanity account, including name, email, provider and the MCP authentication method in use.
list_organizationsreadLists all organizations the user has access to.
list_projectsreadLists all Sanity projects associated with the account.
get_project_studiosreadRetrieves all Studio applications linked to a specific project.
cors_origins_listreadLists all CORS origins configured for a project.
create_projecthigh riskCreates a new Sanity project and initializes it with a dataset and API tokens.
add_cors_originhigh riskAdds CORS origins so client-side requests from those origins reach the project.
cors_origins_deletehigh riskDeletes a CORS origin from a project, which can break a running front end.
get_schemareadFetches a deployed schema, resolving each workspace in order of MCP-managed, Studio-deployed, then legacy.
list_workspace_schemasreadLists every deployed schema for a project and dataset, grouped by source, each with a schema ID for exact reads.
deploy_schemahigh riskDeploys schema types directly to the cloud, changing how content is validated and presented.
deploy_studiohigh riskDeploys a managed Sanity Studio bound to an MCP-managed schema, returning a hosted Studio URL. Requires an existing MCP-managed schema at the same address.
list_datasetsreadLists all datasets in a project.
create_datasetwriteCreates a new dataset with a name and access settings.
update_datasethigh riskModifies a dataset's name or access control settings, which determine whether it is publicly readable.
generate_imagewriteTriggers async AI image generation for a document field. Consumes AI credits.
transform_imagewriteTriggers async AI transformation of an existing image. Consumes AI credits.
list_embeddings_indicesreadLists all available embeddings indices for a dataset.
semantic_searchreadPerforms a semantic search against an embeddings index.
search_docsreadSearches the Sanity documentation.
read_docsreadFetches a specific documentation article.
list_sanity_rulesreadLists the available best-practice development rules.
get_sanity_rulesreadLoads specific best-practice development rules, such as the GROQ rules.
give_sanity_feedbackwriteSubmits feedback about Sanity when the agent hits an unexpected error, a missing capability or unclear documentation.
run_sanity_clihigh riskRuns a limited subset of Sanity CLI commands and returns their output. Commands run without a shell and cannot access the filesystem, prompt for input, run in the background or change authentication.

What Sanity MCP exposes

Sanity MCP groups its 54 tools into 10 categories, each with a suggested starting policy.

Category	Covers	Access	Suggested policy
Documents	GROQ queries, direct ID lookups, creating drafts and release versions, patching documents, publishing, unpublishing, discarding drafts and versions, and creating and listing releases	Read / Write	Edits are saved to the draft or release version, never directly to published content. Scheduling, publishing, archiving and deleting a release are deliberately not available here and must be done in Studio or the HTTP Actions API
Media Library	Listing libraries, their schema, GROQ queries and ID lookups, creating collections, patching the safe subset of asset metadata, publishing and discarding asset drafts, listing what references an asset and linking assets into project documents	Read / Write	media_library_patch_documents is deliberately narrow, but cdnAccessPolicy is published-only and takes effect immediately, so it bypasses the draft step the rest of the surface relies on
Assets & Uploads	Uploading images, video and files to a dataset or Media Library from a public URL, a one-use upload URL, a local file or a ChatGPT web attachment, plus upload status polling and CLI guidance	Read / Write	The from_url tools make Sanity fetch an arbitrary public HTTPS URL on the agent's behalf; pin the allowed sources. File uploads are capped at 50 MiB
Projects & Access	Organizations, projects, studio applications, creating a project, listing, adding and deleting CORS origins, and identifying the authenticated account	Read / Write	create_project initializes a project with its own dataset and API tokens. The CORS tools change which browsers can reach the project, so they are a security control rather than a convenience
Docs & Rules	Searching and reading Sanity documentation, listing and loading best-practice development rules, and submitting product feedback	Read / Write	Account-independent; safe to allow broadly
Schema & Studio	Fetching a deployed schema, listing every deployed schema by source, deploying schema types to the cloud and deploying a managed Studio bound to an MCP-managed schema	Read / Write	deploy_schema changes how content is validated and presented for everyone; deploy_studio creates a publicly reachable hosted Studio URL
Datasets	Listing datasets, creating one with access settings and modifying a dataset's name or access control	Read / Write	update_dataset can change access control settings, which is what decides whether a dataset is public
AI Images	Triggering async AI image generation for a document field and async AI transformation of an existing image	Read / Write	The only two tools that consume AI credits; Sanity suggests disabling them in the client if credit usage is a concern
Search	Listing the embeddings indices available for a dataset and running a semantic search against one	Read	Read-only
CLI	Running a limited subset of Sanity CLI commands and returning their output	Read / Write	A meta-tool: the tool name says nothing about the command run. Commands execute without a shell and cannot touch the filesystem or change authentication, but the reachable surface is the CLI's, not this tool list's