Support for 2025-06-18 MCP spec

[awestwell]

Title:
Add native elicitation/create support to Spring AI MCP Server for MCP 2025-06-18 spec

Description:
The MCP specification (2025-06-18) introduces the Elicitation feature, which allows an MCP server to send an elicitation/create request to the connected client with a flat JSON Schema. The client renders a form in the UI and returns structured user input (accept, decline, or cancel).

This enables scenarios like:

{
  "type": "object",
  "properties": {
    "name": { "type": "string", "title": "Your name", "minLength": 1 }
  },
  "required": ["name"]
}

A tool could call this mid-execution, prompting the client to display a text field and submit button.

Current status in Spring AI:

• The Spring AI MCP Client Boot Starter documentation (link to client docs section) references elicitation in the context of client capabilities.
• The Spring AI MCP Server Boot Starter does not currently expose a native, first-class API for sending elicitation/create requests.
• Implementing elicitation today requires manually retrieving McpSyncServerExchange from ToolContext and issuing exchange.request("elicitation/create", …) with a raw schema map.

Feature request:
Please add a native API to Spring AI MCP Server Boot Starter to support Elicitation end-to-end:

1. Capability detection – At tool runtime, automatically detect if the connected client supports elicitation (advertised in initialize).
2. First-class server API – Provide a simple, type-safe way to send elicitation/create with a provided flat JSON Schema and optional progress token.
3. Response handling – Return a structured result object encapsulating accept / decline / cancel and typed content.
4. Optional annotation/config support – Allow tool methods to declare elicitation prompts directly in @Tool metadata, so Spring can handle sending and parsing the request automatically.

Benefits:

• Keeps elicitation fully within Spring AI idioms (no low-level JSON-RPC calls).
• Improves developer productivity for interactive workflows.
• Enables richer integrations with MCP clients that can render structured forms.

References:

• MCP Spec 2025-06-18 – Elicitation section: modelcontextprotocol.io/spec#elicitation
• Spring AI MCP Server Boot Starter docs: docs.spring.io/spring-ai/reference/api/mcp/mcp-server-boot-starter-docs.html
• Related client-side doc reference (for capability advertisement): client boot starter docs