create-agent-with-sanity-context

To build an AI chatbot with structured Sanity content, use Agent Context to provide schema-aware access, allowing your LLM to query real fields and relationships instead of relying on plain text similarity. Configure an Agent Context document in Sanity Studio with your credentials and MCP URL.

Schema-aware access improves AI agent results by letting the model query Sanity using actual fields and relationships. This prevents vague responses and reduces the extra engineering work typically required when AI lacks structured context from your headless CMS.

You can integrate a Sanity AI assistant with Next.js, SvelteKit, Express, or Python. The Agent Context setup works with any LLM provider that supports tool calling, automatically discovering MCP tools like groq_query and schema_explorer for flexible deployment.

To set up an AI agent with Sanity Agent Context, you need a deployed Sanity Studio, an Agent Context document configured with proper credentials, an MCP URL, and an LLM provider API key to enable tool calling and semantic search over your data.

Available Sanity MCP tools for AI agent integration include initial_context for setup, groq_query for executing structured queries, and schema_explorer for navigating your content structure. These tools are automatically discovered when you configure the Agent Context document.

AI agents return vague results from Sanity data when they lack schema-aware access and rely on plain text similarity. By integrating Agent Context, your agent queries actual fields and relationships, eliminating vague outputs and reducing manual engineering work.