**Source URL:** https://limited.veevavault.dev/medical/mcp/vault-mcp-server/overview # Overview The Vault MCP Server allows Model Context Protocol (MCP) client applications to invoke [Vault AI agent actions](/medical/ai-agents/agent-actions/) on documents and Vault object records. You can invoke these agent actions from within your existing AI client workflows without having to switch to the Vault UI. The Vault MCP Server can integrate directly with client applications, such as Claude, Gemini, and Copilot, as well as other AI agents that delegate tasks to Vault agents. For example, from within your client you could ask for a summary of a Vault document by ID, ask questions about its content, or translate a page. Organizations can also configure [custom agent actions](https://platform.veevavault.help/en/lr/950644), enabling workflows specific to their Vault configuration, such as triggering review processes or querying object records. Each Vault AI agent action that the authenticated user has permission to execute is exposed as a Vault MCP Server tool. The server does not support MCP resources or prompts. Learn more about [configuring agent tools for Vault AI](https://platform.veevavault.help/en/lr/950647). The Vault MCP Server uses Streamable HTTP transport over a single endpoint, with the `{host}` path parameter representing your Vault's DNS. For example, `myvault.veevavault.com`. ```http POST https://{host}/api/ai/mcp ``` The server accepts JSON-RPC 2.0 messages and returns either a single JSON response or a Server-Sent Event (SSE), depending on the request. --- **Previous:** [Vault MCP Server](/medical/mcp/vault-mcp-server) **Next:** [Getting Started](/medical/mcp/vault-mcp-server/getting-started)