Context and KBs now power the Ask AI agent, plus new guides and improvements across the platform

This month we added new guides for dataset environments, upgrading older @sanity/client versions, and diagnosing API request spikes. The docs site also adds an improved “Ask agent” feature, powered by Sanity Context and Knowledge Bases.

• New guide: Development, staging, and production environments explains how drafts, releases, preview, separate datasets, Cloud Clone, and export and import fit together across environments.
• How to use Hot Swapping for datasets now lists what doesn't work through a dataset alias, and scopes hot swapping to recovery and breaking migrations.
• Drafts and versions (previously Drafts) now covers both draft and version documents.
• Backups clarifies which plans include backups, that you turn them on per dataset, and how to confirm they're running.
• Technical limits adds the limit of 200 active releases per dataset, and clarifies that document limits are plan quotas counted across datasets.

• New guide: Archive documents without deleting them compares unpublishing, an archived field, deleting and restoring, and unpublishing in a release.
• New guide: Reorganize your structure without breaking links shows how to rename and move Structure Builder items without breaking saved URLs.
• New guide: Re-enable browser autocomplete on Studio fields covers per-field and studio-wide overrides.
• Validation shows how to restrict publishing by user role with context.currentUser.
• Handle intents in the Structure Tool explains how to route documents of the same type to different lists based on a field value.

• New guide: Upgrade @sanity/client from an older major version covers the breaking changes in each major version from v3 to v8.
• Advanced client patterns with @sanity/client adds error handling and routing requests through a proxy. Working with assets in @sanity/client adds upload progress tracking.
• New guide: Diagnose an API request spike helps you find the cause of unexpected API usage with your request logs.
• Add live content to your application covers private datasets, and the new Relay live draft updates through your server guide shows how to get draft-aware live updates without a token in the browser.
• Command Line Interface (CLI) now lists the CLI's requirements, including the hosts to allowlist on restricted networks.

• New guide: Update an asset to a new version with the API.
• New guide: Use Media Library with AI agents covers the Media Library tools in the Sanity MCP server, including uploading files from your computer.
• Working with video shows how to get static MP4 renditions of Media Library videos.
• Assets and Keeping your data safe clarify that dataset assets are always public on the asset CDN, even in private datasets.

• New guide: Manage stacks covers listing, inspecting, and removing Blueprints stacks. The stack limit is now correctly listed as 10 per scope.
• Add environment variables to functions documents the 4KB size limit, and Functions cheat sheet shows how to keep functions from running during an import or sync.
• Create a PubSub function explains that PubSub functions only run when another function invokes them.
• App SDK deployment shows how to turn on auto-updates for App SDK apps, and warns against putting secrets in SANITY_APP_ environment variables.
• App SDK and TypeGen is rewritten for the typegen.resources workflow, with a new Migrate from experimental TypeGen in the App SDK guide.

• Context MCP tools adds the knowledge_base_search tool. The Context MCP no longer reads a tools URL parameter, so Context MCP reference and Inline initial context into your system prompt now cover filtering tools on the client side.
• Content Agent for Linear is now in the Content Agent navigation, and the Slack and Linear guides explain when the agent can publish.
• Get started with Sanity Content Agent adds the registration step for self-hosted studios.

• New: Spending alerts explains how to get notified as your organization's overage spending grows.
• New: Export activity to Google Cloud Storage and the Activity Feed Export data reference.
• Set up single sign-on with SAML explains that role mapping is additive, applies at first login, and doesn't remove users who leave your identity provider.
• Document mutation patterns and Build a custom role with the Access API explain that purging history requires the editHistory permission, which no predefined role grants.

The Astro docs now recommend the Sanity-maintained @portabletext/astro package (a fork of the community package), and Configuring @sanity/astro documents the clients option added in @sanity/astro 3.6.0.

The docs site's AI assistant is now powered by Sanity Context and Knowledge Bases, the same tools we document for building your own agents.

• Open it from the “Ask agent” button.
• Use Ask about this page on any article to ask a question with that page as context.
• Type a question into search and send it to the Ask agent.

• Code blocks have a Copy link button, so you can link straight to an example.
• Deep links to a code block or heading that is otherwise not rendered (hidden inside a tab, collapsed inside a snippet, etc) will automatically show and render on load.
• We now check the docs for broken internal and external links on a schedule, and block redirects that would break existing pages.