add: elicitation support #535

QuantGeekDev · 2025-06-19T15:51:35Z

Added support for elicitation introduced in the 2025-06-18 specification update. Works across Tools, Resources and Prompts.

This PR depends on #532 for using elicitation from the sdk

Motivation and Context

Adds the ability for users to debug and test the new elicitation feature

How Has This Been Tested?

Here is a video showcasing the functionality and testing

Tested against a demo server I built to test elicitations mcp-elicitations server that implements elicitation in tools, prompts, and resources with the @modelcontextprotocol/sdk@0.1.13
The server is also accessible via this url https://weekly-mink.mcpstudio.io/mcp
You can replicate results by checking out this branch, and connecting the inspector to the hosted server url via streamable-http. Then try to use any of the prompts, tools, or resources. I recommend increasing the request timeout in the inspector configuration

I would appreciate feedback on this - are modals a good way to represent elicitation in the inspector?

Breaking Changes

No, this is a new feature of the protocol that has to be specifically invoked in server code

Types of changes

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to change)
Documentation update

Checklist

I have read the MCP Documentation
My code follows the repository's style guidelines
New and existing tests pass locally
I have added appropriate error handling
I have added or updated documentation as needed

oso-jukumari · 2025-06-21T04:36:04Z

@QuantGeekDev we added your logic to MCPJam/inspector#88 to support elicitation on our client as well. Thank you for helping everyone with your ideas! 🎉

olaservo · 2025-06-28T04:01:48Z

Hi @QuantGeekDev - we finally merged in some fixes to the input forms that were pending for a while, but unfortunately that led to some conflicts in this branch. Let me know if you want me to try resolving them, would definitely like to get your PR in soon.

QuantGeekDev · 2025-06-28T12:12:37Z

Hi @QuantGeekDev - we finally merged in some fixes to the input forms that were pending for a while, but unfortunately that led to some conflicts in this branch.z Let me know if you want me to try resolving them, would definitely like to get your PR in soon.

Thanks for the heads up. I will get this fixed up

QuantGeekDev · 2025-06-28T12:14:12Z

@QuantGeekDev we added your logic to MCPJam/inspector#88 to support elicitation on our client as well. Thank you for helping everyone with your ideas! 🎉

That is awesome :) I hope it was useful - I like your inspector

QuantGeekDev · 2025-07-02T17:55:47Z

Hasn't slipped my mind, just been a bit busy. Should be done within next 3 days

cliffhall · 2025-07-02T21:45:32Z

I would appreciate feedback on this - are modals a good way to represent elicitation in the inspector?

The way you've implemented this looks appropriate. As a modal, we're indicating that the user needs to take an action, not merely that they can choose to interact with a new part of the UI or ignore it.

As you get this one's conflicts resolved, I'll take a look at adding elicitations to our Everything server for in-project testing / demonstrating.

Resolved conflicts by adopting the improved form input handling from main: - Standardized JsonSchemaType.required to only accept string arrays (JSON Schema standard) - Adopted enhanced object and array rendering with proper nested support - Improved boolean field rendering (removed extra wrapper div) - Enhanced array handling with add/remove functionality - Simplified required field validation logic This maintains elicitation functionality while benefiting from the form input fixes.

cliffhall · 2025-07-03T16:09:28Z

I would appreciate feedback on this - are modals a good way to represent elicitation in the inspector?

@QuantGeekDev Coming back to this question, and revising my answer. I think that it should be implemented like sampling on a separate tab. The reason being that while your examples are small, some servers might require a larger form than the modal could comfortably handle.

cliffhall · 2025-07-03T18:17:23Z

I now have a PR for an elicitation tool in the everything reference server for testing this functionality.

cliffhall

A few requests:

Bump SDK version to 1.14.0
Replace elicitation reject action with decline (last minute change to spec doc didn't make it in to schema so it was reverted in SDK version 1.14.0)
Fix required type in jsonUtils.ts (boolean vs string[] issue was merged last week, your branch is older)
Fix use of enumNames which did not make it into the JSON Schema spec and the recommended implementation of labeled enums is with const and oneOf or anyOf, e.g.,

{ "type": "string", "oneOf": [ { "const": "red", "title": "Stop" }, { "const": "amber", "title": "Caution" }, { "const": "green", "title": "Go" } ] }

cliffhall · 2025-07-03T19:39:01Z

package.json

@@ -46,7 +46,7 @@
 "@modelcontextprotocol/inspector-cli": "^0.14.3",
 "@modelcontextprotocol/inspector-client": "^0.14.3",
 "@modelcontextprotocol/inspector-server": "^0.14.3",
- "@modelcontextprotocol/sdk": "^1.12.1",
+ "@modelcontextprotocol/sdk": "^1.13.0",


Suggested change

"@modelcontextprotocol/sdk": "^1.13.0",

"@modelcontextprotocol/sdk": "^1.14.0",

cliffhall · 2025-07-03T20:02:01Z

client/src/utils/jsonUtils.ts

 description?: string;
- required?: boolean;
+ required?: boolean | string[];


Should only be array of strings. This was previously fixed. The boolean type here appears to be a regression. It isn't present on main right now, so I assume you are working from an older branch, and this is one of the conflicts to be resolved.

Indeed from older branch, moved to string array only

cliffhall · 2025-07-03T20:14:45Z

client/src/utils/jsonUtils.ts

+ pattern?: string;
+ format?: string;
+ enum?: string[];
+ enumNames?: string[];


enumNames is not part of the JSONSchema spec. The recommended way to describe enum labels is with oneOf or anyOf.

cliffhall · 2025-07-03T20:15:00Z

client/src/utils/__tests__/jsonUtils.test.ts

+ const schema = {
+ type: "string" as const,
+ enum: ["option1", "option2"],
+ enumNames: ["Option 1", "Option 2"],


enumNames is not part of the JSONSchema spec. The recommended way to describe enum labels is with oneOf or anyOf.

cliffhall · 2025-07-03T20:16:37Z

client/src/utils/__tests__/jsonUtils.test.ts

+ role: {
+ type: "string",
+ enum: ["admin", "user", "guest"],
+ enumNames: ["Administrator", "User", "Guest"],


enumNames is not part of the JSONSchema spec. The recommended way to describe enum labels is with oneOf or anyOf.

cliffhall · 2025-07-03T20:30:33Z

client/src/components/DynamicJsonForm.tsx

@@ -140,36 +140,144 @@ const DynamicJsonForm = ({
 );
 }

+ const isFieldRequired = (fieldPath: string[]): boolean => {
+ if (typeof schema.required === "boolean") {


should only be a string array after fix to jsonUtils.ts.

package-lock.json

cliffhall · 2025-07-03T20:33:38Z

client/src/components/DynamicJsonForm.tsx

+ <option value="">Select an option...</option>
+ {propSchema.enum.map((option, index) => (
+ <option key={option} value={option}>
+ {propSchema.enumNames?.[index] || option}


enumNames is not part of the JSONSchema spec. The recommended way to describe enum labels is with oneOf or anyOf.

cliffhall · 2025-07-03T20:33:49Z

client/src/components/__tests__/DynamicJsonForm.test.tsx

+ const schema: JsonSchemaType = {
+ type: "string",
+ enum: ["val1", "val2"],
+ enumNames: ["Label 1", "Label 2"],


enumNames is not part of the JSONSchema spec. The recommended way to describe enum labels is with oneOf or anyOf.

cliffhall · 2025-07-03T20:33:58Z

client/src/components/__tests__/ElicitationModal.test.tsx

+ country: {
+ type: "string",
+ enum: ["US", "UK", "CA"],
+ enumNames: ["United States", "United Kingdom", "Canada"],


enumNames is not part of the JSONSchema spec. The recommended way to describe enum labels is with oneOf or anyOf.

QuantGeekDev · 2025-07-03T20:44:34Z

A few requests:

Fix required type in jsonUtils.ts

Fix use of enumNames which did not make it into the JSON Schema spec and the recommended use is with const and oneOf or anyOf, e.g.,
{ "type": "string", "oneOf": [ { "const": "red", "title": "Stop" }, { "const": "amber", "title": "Caution" }, { "const": "green", "title": "Go" } ] } 
Revert reject action to decline (last minute change to spec doc didn't make it in to schema so it was reverted in SDK version 1.14.0)

Received, will add this too. Out this weekend but adding it on Monday/Tuesday - does that work out?

Co-authored-by: Cliff Hall <cliff@futurescale.com>

…enums - Make JsonSchemaType.type optional to support const-only schemas - Add JsonSchemaConst type for schema objects using const values - Update oneOf/anyOf to accept both full schemas and const schemas - Add comprehensive tests for oneOf with const and title patterns This fixes TypeScript compilation errors when using the JSON Schema specification's recommended approach for labeled enums: { "type": "string", "oneOf": [{ "const": "red", "title": "Stop" }] } Replaces deprecated enumNames pattern with spec-compliant implementation.

…v/mcp-debug into feature/elicitations

cliffhall · 2025-07-04T17:58:36Z

@QuantGeekDev you need to merge main into your branch. package-lock.json is out of date

QuantGeekDev · 2025-07-04T17:58:53Z

@cliffhall Updated with the feedback, thanks for taking the time to give it. I think this is ready for UAT. Here's a video showing the new functionality in action
Let me know your thoughts

QuantGeekDev · 2025-07-04T18:03:48Z

Part of me thinks it makes sense to redirect the user back to the tool call after they complete the elicitation @cliffhall, you can see that it's a bit awkward navigating back to the tool response

cliffhall · 2025-07-04T18:22:14Z

Part of me thinks it makes sense to redirect the user back to the tool call after they complete the elicitation @cliffhall, you can see that it's a bit awkward navigating back to the tool response

Definitely, we should send the user back to the tool tab to view the response.

olaservo · 2025-07-07T02:37:05Z

Btw, there was a different elicitation demo change open for the Everything server here that was just buried in the other server PRs, I think we're going to use this one and close the other one: modelcontextprotocol/servers#2215

cliffhall · 2025-07-07T16:58:57Z

@QuantGeekDev another thing is, if the user declines elicitation, this current PR is not showing the result after the tool call.

QuantGeekDev added 7 commits June 19, 2025 15:03

feat: add elicitations to useConnection

e163aea

feat: add new schema requirements to DynamicJsonForm

e6166a7

feat: add ElicitationModal

a058229

feat: add elicitation support to jsonUtils

fc939eb

test: new elicitation format for jsonUtils

6c809bd

feat: add elicitation handling to App

124c781

test: elicitation handling

dfbf10c

QuantGeekDev changed the title ~~Feature/elicitations~~ Add Elicitation support Jun 19, 2025

QuantGeekDev changed the title ~~Add Elicitation support~~ add: elicitation support Jun 19, 2025

QuantGeekDev mentioned this pull request Jun 19, 2025

Add support for testing Elicitation capability #524

Open

oso-jukumari mentioned this pull request Jun 21, 2025

Add Elicitation support MCPJam/inspector#88

Merged

gregorym mentioned this pull request Jun 30, 2025

Add Elicitation new standard vercel/mcp-adapter#64

Open

cliffhall mentioned this pull request Jul 3, 2025

Add elicitation support to server-everything modelcontextprotocol/servers#2269

Closed

10 tasks

cliffhall requested changes Jul 3, 2025

View reviewed changes

cliffhall added the waiting on submitter Waiting for the submitter to provide more info label Jul 3, 2025

cliffhall mentioned this pull request Jul 3, 2025

Add enum support #196

Open

9 tasks

QuantGeekDev and others added 4 commits July 4, 2025 09:42

Update package-lock.json

8053e69

Co-authored-by: Cliff Hall <cliff@futurescale.com>

Update client/package.json

ea97f37

Co-authored-by: Cliff Hall <cliff@futurescale.com>

feat: add elicitation request

b4a2a6b

test: Elicitation request

0411383

QuantGeekDev and others added 11 commits July 4, 2025 10:54

feat: add elicitations tab

9539c16

test: elicitations tab

2baf8bc

feat: replace modal logic with new tab request for elicitation

28c1890

refactor: remove dead code

f2166c9

feat: use decline instead of reject

e6c8b94

feat: enum support in dynamic json form

efd5f8a

Merge branch 'feature/elicitations' of https://github.com/QuantGeekDe…

6cddc51

…v/mcp-debug into feature/elicitations

Merge branch 'main' into feature/elicitations

844bbe0

chore: bump sdk version

484b5b8

Merge branch 'feature/elicitations' of https://github.com/QuantGeekDe…

29678dd

…v/mcp-debug into feature/elicitations

olaservo mentioned this pull request Jul 7, 2025

Example showing how to use elicitation modelcontextprotocol/servers#2215

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

add: elicitation support #535

add: elicitation support #535

QuantGeekDev commented Jun 19, 2025 •

edited

Loading

oso-jukumari commented Jun 21, 2025

olaservo commented Jun 28, 2025

QuantGeekDev commented Jun 28, 2025

QuantGeekDev commented Jun 28, 2025

QuantGeekDev commented Jul 2, 2025

cliffhall commented Jul 2, 2025 •

edited

Loading

cliffhall commented Jul 3, 2025 •

edited

Loading

cliffhall commented Jul 3, 2025 •

edited

Loading

cliffhall left a comment •

edited

Loading

cliffhall Jul 3, 2025

cliffhall Jul 3, 2025

QuantGeekDev Jul 4, 2025

cliffhall Jul 3, 2025

cliffhall Jul 3, 2025

cliffhall Jul 3, 2025

cliffhall Jul 3, 2025

Uh oh!

cliffhall Jul 3, 2025

cliffhall Jul 3, 2025

cliffhall Jul 3, 2025

QuantGeekDev commented Jul 3, 2025

cliffhall commented Jul 4, 2025

QuantGeekDev commented Jul 4, 2025 •

edited

Loading

QuantGeekDev commented Jul 4, 2025

cliffhall commented Jul 4, 2025

olaservo commented Jul 7, 2025

cliffhall commented Jul 7, 2025

	"@modelcontextprotocol/sdk": "^1.13.0",
	"@modelcontextprotocol/sdk": "^1.14.0",

add: elicitation support #535

Are you sure you want to change the base?

add: elicitation support #535

Conversation

QuantGeekDev commented Jun 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation and Context

How Has This Been Tested?

Breaking Changes

Types of changes

Checklist

oso-jukumari commented Jun 21, 2025

olaservo commented Jun 28, 2025

QuantGeekDev commented Jun 28, 2025

QuantGeekDev commented Jun 28, 2025

QuantGeekDev commented Jul 2, 2025

cliffhall commented Jul 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

cliffhall commented Jul 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

cliffhall commented Jul 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

cliffhall left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

QuantGeekDev commented Jul 3, 2025

cliffhall commented Jul 4, 2025

QuantGeekDev commented Jul 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

QuantGeekDev commented Jul 4, 2025

cliffhall commented Jul 4, 2025

olaservo commented Jul 7, 2025

cliffhall commented Jul 7, 2025

QuantGeekDev commented Jun 19, 2025 •

edited

Loading

cliffhall commented Jul 2, 2025 •

edited

Loading

cliffhall commented Jul 3, 2025 •

edited

Loading

cliffhall commented Jul 3, 2025 •

edited

Loading

cliffhall left a comment •

edited

Loading

QuantGeekDev commented Jul 4, 2025 •

edited

Loading