mcp-go-sdk
Language:
Go
Stars:
1
Forks:
1
MCP Go SDK
This SDK provides a Go implementation of the Model Context Protocol (MCP), enabling bidirectional communication between clients and servers for tool execution, resource access, and prompt handling.
Features
-
Transport Layer
- Multiple transport options (stdio, SSE, WebSocket)
- Bidirectional communication
- Configurable endpoints and settings
-
Server Implementation
- Tool registration and execution
- Resource pattern matching and access
- Prompt template rendering
- Session management
- Reflection-based handler invocation
-
Core Protocol Types
- JSON-RPC message handling
- MCP-specific types (tools, resources, prompts)
- Capabilities and initialization
Installation
go get github.com/SetiabudiResearch/mcp-go-sdk
Quick Start
Here's a simple example of creating an MCP server:
package main
import (
"context"
"log"
"github.com/SetiabudiResearch/mcp-go-sdk/pkg/mcp/server"
"github.com/SetiabudiResearch/mcp-go-sdk/pkg/mcp/transport"
)
func main() {
// Create a new server
srv := server.NewServer("Example Server")
// Add a tool
srv.AddTool("greet", func(name string) string {
return "Hello, " + name + "!"
}, "Greet a person")
// Create a session
session := server.NewSession(context.Background(), srv)
// Create and start a transport
t := transport.NewStdioTransport(session)
if err := t.Start(); err != nil {
log.Fatal(err)
}
}
Usage Guide
Creating a Server
// Create a new server with a name
srv := server.NewServer("My Server")
// Optionally configure server capabilities
srv.WithCapabilities(protocol.ServerCapabilities{
SupportsAsync: true,
})
Adding Tools
// Add a synchronous tool
srv.AddTool("myTool", func(arg1 string, arg2 int) (string, error) {
return fmt.Sprintf("Processed %s with %d", arg1, arg2), nil
}, "Tool description")
// Add an asynchronous tool
srv.AddAsyncTool("longRunningTool", func(params string) error {
// Long-running operation
return nil
}, "Async tool description")
Adding Resources
// Add a resource with pattern matching
srv.AddResource("files/{path}", func(path string) ([]byte, error) {
return ioutil.ReadFile(path)
}, "Access files")
// Resource patterns support multiple parameters
srv.AddResource("api/{version}/{endpoint}", func(version, endpoint string) (interface{}, error) {
return callAPI(version, endpoint)
}, "API access")
Adding Prompts
// Add a simple text prompt
srv.AddPrompt("confirm", func(action string) string {
return fmt.Sprintf("Are you sure you want to %s?", action)
}, "Confirmation prompt")
// Add a multi-message prompt
srv.AddPrompt("chat", func(context string) []protocol.PromptMessage {
return []protocol.PromptMessage{
{
Role: protocol.RoleAssistant,
Content: protocol.TextContent{
Type: "text",
Text: "How can I help you with " + context + "?",
},
},
}
}, "Chat prompt")
Transport Configuration
// Create a session
session := server.NewSession(context.Background(), srv)
// Stdio transport (for CLI applications)
t := transport.NewStdioTransport(session)
// WebSocket transport (for web applications)
t := transport.NewWebSocketTransport(session, transport.WithAddress(":8080"))
// Server-Sent Events transport (for web browsers)
t := transport.NewSSETransport(session, transport.WithAddress(":8080"))
// Start the transport
if err := t.Start(); err != nil {
log.Fatal(err)
}
Example Applications
See the examples directory for complete example applications:
- File Server: A complete file server implementation using MCP
- Calculator: A simple calculator service
- Chat Bot: An example chat bot using prompts
Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
License
This project is licensed under the MIT License - see the LICENSE file for details.
Publisher info
Setiabudi Research
-- accelerate
More MCP servers built with Go
mcp-kubernetes-ro is a Model Context Protocol (MCP) server providing read-only access to Kubernetes clusters for AI assistants. It enables AI models to list resources, get resource details, retrieve pod logs, discover API resources, and perform base64 encoding/decoding operations - all while maintaining security through read-only access.
A CLI for interacting with GitKraken APIs. Includes an MCP server via gk mcp that not only wraps GitKraken APIs, but also Jira, GitHub, GitLab, and more.
An MCP (Model Context Protocol) server that provides tools for querying OCI registries and image references.