modelcontextprotocol / modelcontextprotocol/java-sdk

Initialization errors in McpAsyncClient not emitted immediately

Open
#393 3 comments 3 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

needs confirmation waiting for user
Dominant language
Java
Stars
3.7k
Forks
1.1k
Avg merge
1d 15h
Merged PRs (30d)
9

Description

Bug Description
When subscribing to exceptions from McpAsyncClient.initialize() using onErrorResumein order to catch initialization failures immediately, nothing is emitted. I suspect this is caused by McpAsyncClient not propagating errors downstream.

I’m new to Reactor and don’t know how to fix this. Any simple tips or pointers, like which class or method handles error propagation, would be really helpful so I can send a PR.

Also, I believe this issue may be related to:

Environment

JDK: zulu-21
gradle: 8.10.2
IntelliJ IDEA 2025.1.1.1 (Community Edition) (K2 mode)

io.modelcontextprotocol java-sdk commit: https://github.com/modelcontextprotocol/java-sdk/commit/8a2f97f1b1995880c492159dd1b4c50266d0b579

Steps to reproduce

  1. Clone and install the repo, checkout https://github.com/modelcontextprotocol/java-sdk/commit/8a2f97f1b1995880c492159dd1b4c50266d0b579
  2. Run the script provided below
  3. Observe that no immediate exception is thrown

Expected behavior

  • When I initialize McpClient with an incorrect transport
  • And internally, the reactor.core.publisher.Operators logger reports:
    java.lang.RuntimeException: Failed to start process with command: [a, a]
  • Then I expect the initialization to fail immediately - not after my configured timeout

Minimal Complete Reproducible example
Main.kt:

import io.modelcontextprotocol.client.transport.StdioClientTransport
import io.modelcontextprotocol.client.transport.ServerParameters
import kotlinx.coroutines.reactive.awaitSingle
import mu.KotlinLogging
import reactor.core.publisher.Mono
import java.time.Duration


private val logger = KotlinLogging.logger {}

suspend fun main() {
    val params = ServerParameters.builder("a").args("b").build()

    val transport = StdioClientTransport(params)

    val client = io.modelcontextprotocol.client.McpClient
        .async(transport)
        .requestTimeout(Duration.ofSeconds(6))
        .initializationTimeout(Duration.ofSeconds(5))
        .build()

    client.initialize()
        .doOnError { t -> logger.error { t }; return@doOnError }
        .onErrorResume { return@onErrorResume Mono.error(it) }
        .awaitSingle()
}

build.gradle.kts:

plugins {
    kotlin("jvm") version "2.1.20"
}

group = "com.anything.mcp"
version = "1.0-SNAPSHOT"

repositories {
    mavenLocal()
    mavenCentral()
}

dependencies {
    implementation("io.modelcontextprotocol.sdk:mcp:0.11.0-SNAPSHOT")
    implementation("io.modelcontextprotocol.sdk:mcp-spring-webflux:0.11.0-SNAPSHOT")
    implementation("io.github.microutils:kotlin-logging:3.0.5")
    implementation("org.jetbrains.kotlinx:kotlinx-coroutines-reactor:1.8.0")
}

kotlin {
    jvmToolchain(21)
}

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by tracing McpAsyncClient.initialize() and the StdioClientTransport startup path used by Main.kt, checking how the process-start failure reaches the Reactor publisher. Use the provided reproduction and the Operators logger output to confirm the failure is emitted immediately rather than only after initializationTimeout; done means the error reaches doOnError/onErrorResume without waiting for the timeout.

Written by the indexing model from the issue text.

Assessment

Tech stack
java, kotlin
Domain
api, backend
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.