Use This MCP client To

Integrate MCP protocol in Rust-based AI applications Enable async tool execution in LLM workflows Manage resources and capabilities in model interactions Handle prompts and responses with structured context Build custom clients for MCP-compatible servers

README

MCP Protocol Client - Very Much a WIP

Rust client implementation for the Model Context Protocol - a structured communication protocol between language models and external tools.

Features

Client implementation to be used in rust projects that want to use the MCP protocol
Async/await support with tokio
Capability negotiation
Tool execution
Resource management
Prompt handling

Usage

use mcp_client_rs::{Protocol, ClientError};

#[tokio::main]
async fn main() -> Result<(), ClientError> {
    let client = Protocol::new(
        "0",  // Protocol version
        "npx", // Command
        vec!["-y", "@modelcontextprotocol/server-github"], // Args
        std::collections::HashMap::new(), // Environment variables
    ).await?;

    // Call tools
    let response = client
        .call_tool(
            "search_repositories",
            serde_json::json!({
                "query": "rust language:rust"
            }),
        )
        .await?;

    println!("{:?}", response);
    Ok(())
}

Installation

Add to your Cargo.toml:

[dependencies]
mcp-client-rs = { git = "https://github.com/tim-schultz/mcp-client-rs.git" }

License

MIT

mcp-client-rs FAQ

How do I install mcp-client-rs in my Rust project?

Add mcp-client-rs as a dependency in your Cargo.toml and use async/await with Tokio runtime.

Does mcp-client-rs support asynchronous operations?

Yes, it fully supports async/await using the Tokio runtime for efficient concurrency.

Can mcp-client-rs negotiate capabilities with MCP servers?

Yes, it includes capability negotiation to ensure compatible interactions with MCP servers.

How does mcp-client-rs handle tool execution?

It provides methods to call and execute tools exposed by MCP servers with structured JSON parameters.

Is mcp-client-rs limited to Anthropic models?

No, while developed with Anthropic's MCP in mind, it is protocol-compliant and can work with any MCP-compatible server, including those for OpenAI, Claude, and Gemini.

What kind of resource management does mcp-client-rs offer?

It manages resources such as tool handles and session state to maintain efficient and consistent model interactions.

Can I use mcp-client-rs for prompt handling?

Yes, it supports prompt management to structure and send prompts within the MCP framework.

Is mcp-client-rs production-ready?

It is currently a work in progress (WIP), so it may lack some features or stability for production use.