MCP elicitation: how agents ask users for input mid-tool-call

MCP elicitation enables an MCP server to request missing user input or confirmation during tool calls, prompts, or resource reads rather than forcing agents to gather every detail upfront or make unsafe guesses. Using a multi-round-trip pattern, the server returns an input-required result with a form schema or URL request, the client presents it to the user, and then retries the original request with accepted responses or reports a decline or cancellation. Form mode supports limited, flat JSON Schema-based fields for non-sensitive information, while URL mode directs users to an external page for OAuth, secrets, and other protected steps that must remain outside the client and model context. Customer-facing AI products must declare supported elicitation capabilities, connect MCP request handlers to their interface, securely manage pending requests, and validate responses server-side. Elicitation is best used for user-only information, consequential in-tool confirmations, and ambiguous choices, while known values should remain tool arguments, general tool execution approval should use client-side gates, and credentials should use URL-based authorization or dedicated connection flows such as Nango.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.