MCP Elicitation is a mechanism in the Model Context Protocol (MCP) for an MCP server to request additional input from the user through the client. It is used when a server-side operation cannot proceed safely or correctly without user-provided information.
Examples include asking the user to choose an account, confirm a missing date range, provide a required form field, or approve opening an external authorization URL.
MCP Elicitation is important because tool calls are not always one-shot API operations. A server may discover during execution that it needs structured user input. The client must present that request clearly, bind it to the requesting server, and allow the user to accept, decline, or cancel.
Security considerations include:
- showing which server is requesting information;
- explaining why the information is needed;
- avoiding sensitive data in URLs;
- requiring consent before opening external links;
- not allowing the model to silently fill user-only inputs; and
- binding responses to the correct user and client session.
In the 2026 stateless MCP direction, elicitation moves away from push-style server-to-client JSON-RPC requests and toward in-band inputRequired style results. This aligns with the MCP Stateless Protocol by making user-input requirements part of the original operation's response and retry flow.
MCP Elicitation is not the same as ordinary tool calling. Tool calling is model-to-application action selection. Elicitation is server-to-client user-input negotiation within the MCP protocol.
The MCP specification documents elicitation security requirements in its elicitation draft.
The LLM Knowledge Base is a collection of bite-sized explanations for commonly used terms and abbreviations related to Large Language Models and Generative AI.
It's an educational resource that helps you stay up-to-date with the latest developments in AI research and its applications.