- add a simple streamable http server example with auth - update readme: replace quickstart example from sse server to streamable server - add readme files for samples, simple-streamable-server, notebooks closes #170 ## How Has This Been Tested? knit and inspector ## Breaking Changes NaN ## 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) - [x] Documentation update ## Checklist - [x] I have read the [MCP Documentation](https://modelcontextprotocol.io) - [x] My code follows the repository's style guidelines - [x] New and existing tests pass locally - [ ] I have added appropriate error handling - [x] I have added or updated documentation as needed
1.6 KiB
1.6 KiB
Kotlin MCP Client
An interactive CLI client that connects to any MCP server over STDIO and pipes queries through Anthropic's Claude API.
Overview
This sample demonstrates a complete MCP client workflow: launching an MCP server as a subprocess, discovering its tools, converting them to Anthropic's tool format, and running an interactive chat loop where Claude can call server tools on behalf of the user.
Prerequisites
- JDK 17+
- An
ANTHROPIC_API_KEYenvironment variable set with a valid Anthropic API key - An MCP server script to connect to (
.js,.py, or.jar)
Build & Run
Run the client, passing the path to an MCP server:
# Connect to a JVM server
./gradlew run --args="path/to/server.jar"
# Connect to a Python server
./gradlew run --args="path/to/server.py"
# Connect to a Node.js server
./gradlew run --args="path/to/build/index.js"
Note
The client uses STDIO transport, so it launches the MCP server as a subprocess. Ensure the server script is executable and is a valid
.js,.py, or.jarfile.
MCP Capabilities
From the client perspective, this sample demonstrates:
- Tool discovery — lists tools from the connected server and converts them to Anthropic's tool format.
- Tool invocation — when Claude's response requests a tool call, the client invokes the corresponding MCP tool and feeds the result back into the conversation.