CallToolResult spec issue..

[stallent]

@mattt Figured out the issue. The code is not spec compliant for CallToolResult outlined here:

https://github.com/modelcontextprotocol/specification/blob/main/schema/draft/schema.json

In your code CallTool.Result.isError should be an optional. Happy to create a PR later today if you'd like.

-stephen

[mattt]

@stallent Thanks for investigating! Yeah, this is interesting.

Taking a look at the spec for CallToolResult:

CallToolResult description:
The server's response to a tool call.

Any errors that originate from the tool SHOULD be reported inside the result object, with isError set to true, not as an MCP protocol-level error response. Otherwise, the LLM would not be able to see that an error occurred and self-correct.

However, any errors in finding the tool, an error indicating that the server does not support tool calls, or any other exceptional conditions, should be reported as an MCP error response.

isError description:
Whether the tool call ended in an error.
If not set, this is assumed to be false (the call was successful).

It then lists content as the only required field.

So my reading of this is that sending "isError": false" should be valid, albeit redundant. But I think you're right that the property should be optional and sent only if true.

I just opened #11 making this change.

[mattt]

This change is now available in 0.3.2

[stallent]

@mattt the issue was when hitting an MCPServer that didn't return an error. This code works fine if its on both sides because then spec adherence isn't an issue. But we are hitting many different ones and thats where the casting was failing. Thank you for the prompt fix!

[mattt]

@stallent Ahhhh, now I get it. Thanks for clarifying. Yeah, that makes sense.

Thanks again for reporting this issue. If you bump on anything else, please do open an issue.