Overview

Elicitations allow MCP servers to request structured input from users during tool execution. This enables interactive workflows where servers need additional information mid-operation.

Key Capabilities

🔔 Server-initiated modal

Server-initiated modal dialogs during tool execution

📝 Supports multiple types

Support for Boolean, Number, String, and Enum input types

📅 mart input fields

Smart input fields (email, date, datetime, URI validation)

📋 Dedicated Elicitations tab

Dedicated Elicitations tab for pending requests

📜 Complete history

Complete history with status tracking

⏱️ Configurable timeout

Configurable timeout behavior (optional)

✅ Accept/Reject workflow

User can Accept/Reject to determine process

How Elicitations Work

Typical Flow

  1. User calls a tool (from Tools tab or via Chat)
  2. Server needs input (e.g., “Approve this action?”)
  3. Modal appears with the server’s message and input fields
  4. User responds by filling fields and clicking Accept/Reject
  5. Server receives response and continues execution

Server-Initiated Requests

When a server requests input during execution, a modal automatically appears.

Modal contents:

  • Server’s message/question
  • Input fields based on requested schema
  • Accept and Reject buttons

📸 Screenshot:

elicitation-modal-dialog

Description: Show the elicitation modal with a server message and input fields visible

Input Types

Boolean

Display: Toggle switch with Yes/No labels

Use case: Confirmations, approvals

Example: “Approve this operation?”

📸 Screenshot:

elicitation-boolean-toggle

Description: Show a boolean toggle switch in an elicitation form

Number

Display: Numeric input field

Use case: Quantities, thresholds, timeouts

Example: “Enter timeout in seconds”

String

Display: Smart input adapting to format

Formats:

  • email: Email input with validation
  • date: Date picker
  • datetime: DateTime picker
  • uri: URL input with validation
  • text (default): Standard text input

📸 Screenshot:

elicitation-string-formats

Description: Show examples of different string input types (date picker, email input, etc.) in an elicitation form

Enum

Display: Dropdown select with predefined options

Use case: Multiple choice selections

Example: Option (0-10, 11-20)

📸 Screenshot:

elicitation-enum-dropdown

Description: Show an enum dropdown with multiple options in an elicitation form

Elicitations Tab

Navigate to 🤝 Elicitations tab to manage all requests.

Pending Requests

View all requests awaiting your response.

Displayed information:

  • Connection name
  • Server message
  • Required input fields
  • Accept/Reject buttons

How to respond:

  1. Fill in all required fields
  2. Click Accept to submit response
  3. Or click Reject to decline

📸 Screenshot:

elicitations-pending-list

Description: Show the accepting the elicitation to progress it, then we see what values were entered

Screenshot:

elicitation-response

Description: Show the response from the tool after everything has been progressed through

History

Review all past elicitation requests.

Status badges:

  • Pending (yellow): Awaiting response
  • Accepted (green): User accepted and submitted values
  • Rejected (gray): User declined or request timed out

Information shown:

  • Timestamp
  • Connection name
  • Server message
  • Status
  • Submitted values (for accepted requests)

📸 Screenshot:

elicitations-history-section

Description: Show the History section with multiple elicitations displaying different status badges

Configuration

Timeout Settings

Configure timeout behavior in appsettings.json:

  {
  "Elicitation": {
    "TimeoutSeconds": 0
  }
}

Values:

  • 0 (default): No timeout, wait indefinitely
  • Positive number: Timeout after N seconds, auto-reject

Use cases for timeout:

  • Prevent indefinite hangs
  • Enforce response time limits
  • Automated workflows

Common Workflows

Approval During Tool Execution

  1. User calls a tool that requires approval
  2. Modal appears: “Approve deletion of 50 records?”
  3. User reviews and clicks Accept
  4. Tool continues and completes deletion

Multi-Field Data Entry

  1. Tool needs structured input (e.g., user profile)
  2. Modal shows multiple fields (name, email, age)
  3. User fills all fields
  4. Clicks Accept
  5. Server receives complete data object

Optional Rejection

  1. Modal appears with request
  2. User decides not to proceed
  3. Clicks Reject
  4. Server receives rejection response
  5. Tool execution halts or takes alternative path

Troubleshooting

Solutions:

  1. Check that tool execution started
  2. Verify server supports elicitations
  3. Review console for errors
  4. Ensure browser allows modals

Input Validation Fails

Solutions:

  1. Check format requirements (email, URI, etc.)
  2. Fill all required fields
  3. Verify numeric values are valid numbers
  4. Check enum selection is from allowed values

Request Times Out

Solutions:

  1. Respond more quickly
  2. Increase TimeoutSeconds in config
  3. Set TimeoutSeconds: 0 for no timeout
  4. Check if server enforces own timeout

Elicitation Implementation

Here’s a method that implements the elicitation shown in the screenshots above. You can find the original code ➡️ GuessTheNumber

Code extract 

  // uses a static readonly Random which is a better pattern for thread safety and randomness quality.
private static readonly Random _random = Random.Shared;

[McpServerTool]
[Description("A simple game where the user has to guess a number between 1 and 10. #elicitation")]
public async Task<string> GuessTheNumber(McpServer server, CancellationToken token)
{
    // Ask if they want to play
    var playResponse = await server.ElicitAsync(new ElicitRequestParams
    {
        Message = "Do you want to play a game?",
        RequestedSchema = new ElicitRequestParams.RequestSchema
        {
            Properties =
            {
                ["Answer"] = new ElicitRequestParams.BooleanSchema(),
                ["Options"] = new ElicitRequestParams.EnumSchema 
                { 
                    EnumNames = new List<string> { "0-10", "11-20" } 
                },
                ["Title"] = new ElicitRequestParams.StringSchema 
                { 
                    MaxLength = 30, 
                    Description = "The title of the game" 
                },
                ["Date"] = new ElicitRequestParams.StringSchema 
                { 
                    Description = "Enter a date: DD/MM/YYYY", 
                    Format = "date" 
                }
            }
        }
    }, token);

    if (playResponse.Action != "accept" || playResponse.Content?["Answer"].ValueKind != JsonValueKind.True)
        return "Maybe next time!";

    // Get game parameters and ask for guess
    var title = playResponse.Content?["Title"].GetString() ?? "Guess the number";
    var useHighRange = playResponse.Content?["Options"].GetString() == "11-20";
    var (min, max) = useHighRange ? (11, 20) : (0, 10);

    _logger.LogInformation("Game title: {Title}, Range: {Min}-{Max}", title, min, max);

    var guessResponse = await server.ElicitAsync(new ElicitRequestParams
    {
        Message = title,
        RequestedSchema = new ElicitRequestParams.RequestSchema
        {
            Properties =
            {
                ["Answer"] = new ElicitRequestParams.NumberSchema
                {
                    Minimum = min,
                    Maximum = max,
                    Description = $"Enter a value between {min} and {max}"
                }
            }
        }
    }, token);

    var guess = guessResponse.Content?["Answer"].GetInt32();
    var correctAnswer = _random.Next(min, max + 1);

    _logger.LogInformation("Guess: {Guess}, Correct answer: {Answer}", guess, correctAnswer);

    return guess == correctAnswer 
        ? "You guessed correctly!" 
        : $"You guessed wrong! Correct answer was {correctAnswer}";
}

Next Steps