Here’s a familiar problem: You’re prototyping with Claude or Cursor, building features fast, but the moment you need content management, everything slows down. You either hardcode strings (technical debt), reach for markdown files (works until someone else needs to edit), or stop to set up a CMS manually (context switch, momentum lost).
Sanity’s MCP server and Content Agent fix this. Both are now generally available.
The MCP Server: CMS setup without leaving your editor
The Remote MCP Server connects AI coding assistants directly to Sanity. Claude, Cursor, GitHub Copilot, VS Code with Copilot. They can now create schemas, add content, query documents, and manage releases through conversation.
What this looks like in practice:
- “Create a blog schema with title, author reference, and rich text body” → schema deployed
- “Add three sample posts about AI development” → content created
- “Query all posts missing meta descriptions” → GROQ query generated and run
The server includes 40+ operations: document CRUD, schema deployment, content releases, even AI image generation. Authentication uses your existing Sanity login. No API tokens to manage for getting started.
One detail that matters: the server ships with 200+ lines of instructions and skills teaching AI agents how Sanity actually works. These update automatically from Sanity’s official repo. No stale .cursorrules files sitting in your project.
For prototyping, you can go from prompt to working CMS in minutes. For production, it handles content audits, release validation, and debugging (“why isn’t this content showing up?”).
Content Agent: Bulk operations through conversation
Content Agent lives inside the Sanity Dashboard. It knows your schema, your documents, your references. Content teams can talk to it directly.
The use cases that keep coming up:
SEO at scale. One team staged 227 metadata edits from a CSV in a single conversation. Previously: spreadsheet, copy-paste, hours of clicking.
Content audits. “Which product pages are missing descriptions?” Instant answer. No GROQ knowledge required.
Bulk updates. A team needed to add Facebook links to 1,100+ office locations. Two days of work became one conversation.
Research. Content Agent can search the web and cross-reference against your existing content. Find gaps, check facts, pull in context.
The safety model: every edit lands in a review panel before publishing. AI proposes, humans approve. You can also bundle changes into Content Releases for coordinated publishing.
Why this matters for AI Development
If you’re building AI-powered applications, your content infrastructure needs to keep up.
Structured content works better as AI context than page blobs. When your content is typed, referenced, and queryable, AI agents can reason about it. “Find all products in category X that mention feature Y” is a GROQ query, not a regex nightmare.
Real-time sync means AI agents and human editors don’t collide. The Content Lake handles concurrent edits. No locked documents, no merge conflicts.
Programmable workflows mean you can automate content operations the same way you automate deployments. Webhooks, functions, scheduled releases. Content as code.
What teams are seeing
Development teams using the MCP Server during prototyping ship content-enabled features in hours instead of weeks. The CMS isn’t a separate project anymore. It’s part of the build.
Content teams using Content Agent report audit times dropping from days to minutes. The questions that used to require a developer ticket (“how many articles reference this product?”) now get answered in conversation.
Morning Brew, Complex, CoinDesk, Braze, Home Instead. These teams shaped Content Agent during early access. The patterns that emerged: multi-language migrations, bulk metadata updates, onboarding new editors through conversation instead of documentation.
Get started
MCP Server: Run npx sanity@latest mcp configure to connect your AI coding tool. No self-hosting required.
Content Agent: Available now. Look for the agent icon in your dashboard.
Both work with existing Sanity projects. Both are included on all plans with usage-based pricing for AI operations.
Learn more at:
