Trait PromptDefinition

Source

pub trait PromptDefinition:
    HasPromptMetadata
    + HasPromptDescription
    + HasPromptArguments
    + HasPromptAnnotations
    + HasPromptMeta
    + Send
    + Sync {
    // Provided methods
    fn display_name(&self) -> &str { ... }
    fn to_prompt(&self) -> Prompt { ... }
}

Expand description

Complete prompt definition - composed from fine-grained traits Complete MCP Prompt Creation - Build reusable prompt templates that generate contextual content.

This trait represents a complete, working MCP prompt that can be registered with a server and invoked by clients. When you implement the required metadata traits, you automatically get PromptDefinition for free via blanket implementation.

§What This Enables

Prompts implementing PromptDefinition become full MCP citizens that are:

🔍 Discoverable via prompts/list requests
🎯 Executable via prompts/get requests with arguments
✅ Validated against their argument specifications
📝 Template-driven for consistent, reusable content generation

§Complete Working Example

use std::collections::HashMap;

// This struct will automatically implement PromptDefinition!
struct CodeReviewPrompt {
    review_style: String,
}

impl HasPromptMetadata for CodeReviewPrompt {
    fn name(&self) -> &str { "code_review" }
    fn title(&self) -> Option<&str> { Some("AI Code Review Assistant") }
}

impl HasPromptDescription for CodeReviewPrompt {
    fn description(&self) -> Option<&str> {
        Some("Generate detailed, actionable code review comments with security and performance insights")
    }
}

impl HasPromptArguments for CodeReviewPrompt {
    fn arguments(&self) -> Option<&Vec<PromptArgument>> {
        static ARGS: std::sync::OnceLock<Vec<PromptArgument>> = std::sync::OnceLock::new();
        Some(ARGS.get_or_init(|| vec![
            PromptArgument {
                name: "language".to_string(),
                description: Some("Programming language (rust, python, javascript, etc.)".to_string()),
                required: Some(true),
            },
            PromptArgument {
                name: "code".to_string(),
                description: Some("Source code to review".to_string()),
                required: Some(true),
            },
            PromptArgument {
                name: "focus".to_string(),
                description: Some("Review focus: security, performance, style, or general".to_string()),
                required: Some(false),
            },
        ]))
    }
}

// Implement remaining required traits...
impl HasPromptAnnotations for CodeReviewPrompt { fn annotations(&self) -> Option<&crate::meta::Annotations> { None } }
impl HasPromptMeta for CodeReviewPrompt { fn prompt_meta(&self) -> Option<&HashMap<String, serde_json::Value>> { None } }

// 🎉 CodeReviewPrompt now automatically implements PromptDefinition!
// Clients can discover it and generate code reviews with customizable parameters

§Usage Patterns

§Easy: Use Derive Macros

#[derive(McpPrompt)]
#[prompt(name = "doc_generator", description = "Generate API documentation")]
struct DocumentationPrompt { api_spec: String }

§Advanced: Manual Implementation (shown above)

Perfect when you need complex argument validation or dynamic prompt generation.

§Real-World Prompt Ideas

Code Prompts: Code reviews, documentation generation, refactoring suggestions, test creation
Content Prompts: Blog post outlines, email templates, meeting summaries, report generation
Analysis Prompts: Data insights, security assessments, performance analysis, bug reports
Learning Prompts: Tutorial creation, concept explanations, example generation, Q&A formatting
Workflow Prompts: Task breakdowns, project planning, requirement analysis, user stories

§How It Works in MCP

Registration: Server registers your prompt during startup
Discovery: Client calls prompts/list → sees available prompts with arguments
Parameter Collection: Client gathers required arguments from user/context
Generation: Client calls prompts/get with prompt name and arguments
Template Processing: Your prompt generates contextual content
Response: Framework delivers generated content back to client