Content
# D MCP Server Library
A D language implementation of the Model Context Protocol (MCP) server, focusing on stdio transport. This library enables D applications to provide functionality to AI language models through a standardized protocol.
## Features
- MCP protocol v2024-11-05 support
- stdio transport implementation
- Type-safe schema builder for tool parameters
- Resource system with static, dynamic, and template resources
- Prompt system with text, image, and resource content
- Change notifications for resources and prompts
- Clean, non-reflective design with type-safe APIs
## Installation
Add to your `dub.json`:
```json
{
"dependencies": {
"mcp": "~>0.1.0"
}
}
```
## Usage
See the [example application](source/app.d) for a complete working example.
### Basic Server Setup
```d
import mcp.server;
import mcp.schema;
import std.json;
void main() {
// Create server with default stdio transport
auto server = new MCPServer("My MCP Server", "1.0.0");
// Add a tool
server.addTool(
"add", // Tool name
"Add two numbers", // Description
SchemaBuilder.object() // Input schema
.addProperty("a",
SchemaBuilder.number()
.setDescription("First number")
.range(-1000, 1000))
.addProperty("b",
SchemaBuilder.number()
.setDescription("Second number")
.range(-1000, 1000)),
(JSONValue args) { // Tool handler
auto a = args["a"].get!double;
auto b = args["b"].get!double;
return JSONValue(a + b);
}
);
// Start server and begin processing messages
server.start();
}
```
### Resources
```d
import mcp.resources : ResourceContents;
// Static resource
server.addResource(
"memory://greeting", // Resource URI
"Greeting", // Name
"A friendly greeting", // Description
() => ResourceContents.makeText(
"text/plain",
"Hello, World!"
)
);
// Dynamic resource
server.addDynamicResource(
"docs://", // Base URI
"Documentation", // Name
"Access documentation files", // Description
(string path) {
import std.path : buildPath;
import std.file : read, exists;
auto fullPath = buildPath("docs", path);
if (exists(fullPath)) {
return ResourceContents.makeText(
"text/markdown",
cast(string)read(fullPath)
);
}
throw new Exception("File not found: " ~ fullPath);
}
);
// Template resource
server.addTemplate(
"weather://{city}/{date}", // URI template
"Weather Forecast", // Name
"Get weather forecast for a city and date", // Description
"application/json", // MIME type
(string[string] params) {
// Extract parameters from URI
auto city = params["city"];
auto date = params["date"];
// Generate content based on parameters
return ResourceContents.makeText(
"application/json",
`{"city":"` ~ city ~ `","date":"` ~ date ~ `","temp":72}`
);
}
);
```
### Prompts
```d
import mcp.prompts;
// Simple text prompt
server.addPrompt(
"greet", // Prompt name
"A friendly greeting prompt", // Description
[
PromptArgument("name", "User's name", true),
PromptArgument("language", "Language code (en/es)", false)
],
(string name, string[string] args) {
string greeting = "language" in args && args["language"] == "es" ?
"¡Hola" : "Hello";
return PromptResponse(
"Greeting response",
[PromptMessage.text("assistant", greeting ~ " " ~ args["name"] ~ "!")]
);
}
);
// Multi-modal prompt with image
server.addPrompt(
"image_badge",
"Generate a user badge with avatar",
[
PromptArgument("username", "Username to display", true),
PromptArgument("role", "User role (admin/user)", true)
],
(string name, string[string] args) {
// Example base64 encoded image data
auto imageData = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8z8BQDwAEhQGAhKmMIQAAAABJRU5ErkJggg==";
return PromptResponse(
"User badge for " ~ args["username"],
[
PromptMessage.text(
"assistant",
"User Badge for " ~ args["username"] ~ " (" ~ args["role"] ~ ")"
),
PromptMessage.image(
"assistant",
imageData,
"image/png"
)
]
);
}
);
```
### Change Notifications
```d
// Get notifier callback when adding a resource
auto notifyChanged = server.addResource(
"memory://greeting",
"Greeting",
"A friendly greeting",
() => ResourceContents.makeText(
"text/plain",
"Hello, World!"
)
);
// Later, when the resource content changes:
notifyChanged(); // This will trigger a notification to clients
```
### Schema Builder
The schema builder provides a fluent interface for defining JSON schemas:
```d
// Object schema with nested properties
auto personSchema = SchemaBuilder.object()
.addProperty("name",
SchemaBuilder.string_()
.setDescription("User name")
.stringLength(1, 100))
.addProperty("age",
SchemaBuilder.integer()
.setDescription("User age")
.range(0, 150))
.addProperty("email",
SchemaBuilder.string_()
.setPattern(r"^[^@]+@[^@]+\.[^@]+$"))
.addProperty("tags",
SchemaBuilder.array(
SchemaBuilder.string_()
).optional());
// Array schema with length constraints
auto arraySchema = SchemaBuilder.array(
SchemaBuilder.string_()
)
.length(1, 10) // Min and max items
.unique(); // Require unique items
// Enum schema for string values
auto colorSchema = SchemaBuilder.enum_(
"red", "green", "blue"
);
// Numeric schema with range and multiple-of constraints
auto numberSchema = SchemaBuilder.number()
.range(0, 100)
.setMultipleOf(0.5);
```
## Protocol Support
The library implements the MCP specification v2024-11-05:
- JSON-RPC 2.0 message format
- Initialize/initialized lifecycle
- Tool registration and execution with schema validation
- Resource access with static, dynamic, and template resources
- Prompt system with text, image, and resource content
- Change notifications for resources and prompts
## Custom Transports
The library uses stdio transport by default, but you can implement custom transports:
```d
class MyCustomTransport : Transport {
// Implement the Transport interface methods
void setMessageHandler(void delegate(JSONValue) handler) { ... }
void handleMessage(JSONValue message) { ... }
void sendMessage(JSONValue message) { ... }
void run() { ... }
void close() { ... }
}
// Use custom transport
auto transport = new MyCustomTransport();
auto server = new MCPServer(transport, "My Server", "1.0.0");
```
## Building
```bash
# Build library
dub build
# Run example
dub run
# Run tests
dub test
```
## License
MIT License
## Contributing
1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Run tests to ensure everything works
5. Submit a pull request
