MCP Prompt Optimizer

#!/usr/bin/env node /** * MCP Prompt Optimizer Server * * This MCP server implements the Optimized Prompts Framework (OTA Loop) * as a tool that can be called before making AI requests. * * It provides a `optimize_prompt` tool that analyzes and enhances user prompts. */ import { Server } from '@modelcontextprotocol/sdk/server/index.js'; import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'; import { CallToolRequestSchema, ListToolsRequestSchema, Tool, } from '@modelcontextprotocol/sdk/types.js'; import * as fs from 'fs/promises'; import * as path from 'path'; interface PromptAnalysis { domain: string; clarity_score: number; risk_flags: string[]; questions: string[]; optimized_prompt: string; optimization_header: string; assumptions: string[]; needs_clarification: boolean; } class PromptOptimizer { private frameworkPath: string; private framework: string | null = null; constructor(frameworkPath?: string) { this.frameworkPath = frameworkPath || path.join(process.env.HOME || '', 'prompts', 'optimized_prompts.md'); } async loadFramework(): Promise<void> { try { this.framework = await fs.readFile(this.frameworkPath, 'utf-8'); } catch (error) { console.error('Warning: Could not load framework file:', error); this.framework = null; } } /** * Analyze a prompt and determine domain, clarity, and risks */ analyzePrompt(prompt: string): Partial<PromptAnalysis> { const analysis: Partial<PromptAnalysis> = { domain: this.detectDomain(prompt), clarity_score: this.calculateClarityScore(prompt), risk_flags: this.detectRisks(prompt), questions: [], assumptions: [], }; return analysis; } /** * Detect the domain of the request */ private detectDomain(prompt: string): string { const lower = prompt.toLowerCase(); // Code-related keywords if (/(code|function|api|debug|error|implement|build|create.*app|deploy|test)/i.test(prompt)) { return 'code'; } // UX-related keywords if (/(ui|ux|design|interface|user.*experience|accessibility|usability|wireframe)/i.test(prompt)) { return 'UX'; } // Data-related keywords if (/(data|analyze|statistics|metrics|chart|graph|calculate|sql|query)/i.test(prompt)) { return 'data'; } // Writing-related keywords if (/(write|blog|article|content|copy|email|documentation|readme)/i.test(prompt)) { return 'writing'; } // Research-related keywords if (/(research|study|investigate|compare|evaluate|analyze.*market)/i.test(prompt)) { return 'research'; } // Finance-related keywords if (/(roi|revenue|cost|budget|finance|pricing|valuation)/i.test(prompt)) { return 'finance'; } // Product-related keywords if (/(product|feature|roadmap|strategy|market.*plan|gtm)/i.test(prompt)) { return 'product'; } return 'misc'; } /** * Calculate clarity score (0-1) based on presence of key elements */ private calculateClarityScore(prompt: string): number { let score = 0; const weights = { goal: 0.30, context: 0.25, format: 0.15, criteria: 0.20, technical: 0.10, }; // Goal clarity: Has clear objective? if (/\b(create|build|implement|analyze|design|write|calculate|help.*with)\b/i.test(prompt)) { score += weights.goal * 0.8; } if (prompt.length > 50) { score += weights.goal * 0.2; // More detail usually means clearer goal } // Context: Provides background? if (/(using|with|for|in|on|my|our|the)\s+\w+/i.test(prompt)) { score += weights.context * 0.5; } if (prompt.split(/\s+/).length > 20) { score += weights.context * 0.5; // Longer prompts often have more context } // Format: Specifies output format? if (/(code|markdown|json|csv|list|table|document|script|function)/i.test(prompt)) { score += weights.format; } // Criteria: Has success criteria or constraints? if (/(must|should|need|require|ensure|with.*test|accessible|secure)/i.test(prompt)) { score += weights.criteria; } // Technical: Includes technical details? if (/(react|python|node|sql|api|database|framework|library|version)/i.test(prompt)) { score += weights.technical; } return Math.min(score, 1.0); } /** * Detect potential risks in the prompt */ private detectRisks(prompt: string): string[] { const risks: string[] = []; const lower = prompt.toLowerCase(); // Security risks if (/(password|auth|login|token|secret|key|credential|hack|exploit|vulnerability)/i.test(prompt)) { risks.push('security'); } // Privacy risks if (/(email|phone|address|ssn|personal.*data|pii|gdpr)/i.test(prompt)) { risks.push('privacy'); } // Policy violations (basic detection) if (/(fake|bypass|circumvent|illegal|hack.*into|crack|steal)/i.test(prompt)) { risks.push('policy'); } // Safety concerns if (/(harm|dangerous|weapon|explosive|poison|drug)/i.test(prompt)) { risks.push('safety'); } // Compliance/legal if (/(medical.*advice|legal.*advice|financial.*advice|tax|investment.*recommendation)/i.test(prompt)) { risks.push('compliance'); } return risks; } /** * Generate targeted questions based on missing information */ generateQuestions(prompt: string, domain: string, clarityScore: number): string[] { const questions: string[] = []; if (clarityScore >= 0.6) { return questions; // No questions needed } // Domain-specific questions switch (domain) { case 'code': if (!/(react|vue|python|node|java|go|rust|typescript|javascript)/i.test(prompt)) { questions.push('What programming language or framework are you using?'); } if (!/(function|class|component|api|endpoint|feature)/i.test(prompt)) { questions.push('What specific feature or component are you building?'); } if (questions.length < 3 && !/(test|validation|security)/i.test(prompt)) { questions.push('Do you need tests, validation, or specific security considerations?'); } break; case 'UX': if (!/(user|customer|audience|people)/i.test(prompt)) { questions.push('Who are the target users for this interface?'); } if (!/(mobile|desktop|responsive|web|app)/i.test(prompt)) { questions.push('What platform is this for (web, mobile, desktop)?'); } break; case 'data': if (!/(rows|records|dataset|table|csv|json)/i.test(prompt)) { questions.push('What is the shape/structure of your data?'); } if (!/(calculate|sum|average|count|metric)/i.test(prompt)) { questions.push('What specific metrics or calculations do you need?'); } break; case 'writing': if (!/(audience|reader|customer|user|stakeholder)/i.test(prompt)) { questions.push('Who is the target audience?'); } if (!/(\d+\s*(word|page|paragraph|character))/i.test(prompt)) { questions.push('What length are you targeting (word count, pages)?'); } if (questions.length < 3 && !/(formal|casual|technical|friendly|professional)/i.test(prompt)) { questions.push('What tone should this have (formal, casual, technical)?'); } break; case 'product': case 'finance': case 'research': if (prompt.split(/\s+/).length < 15) { questions.push('Can you provide more context about your goal?'); } if (!/(timeline|deadline|when|by|urgent)/i.test(prompt)) { questions.push('What is the timeline or urgency for this?'); } break; default: // Generic questions for misc domain if (prompt.split(/\s+/).length < 10) { questions.push('Can you provide more detail about what you need?'); } if (!/(for|with|using|in)/i.test(prompt)) { questions.push('What is the context or setting for this request?'); } } // Limit to 3 questions return questions.slice(0, 3); } /** * Create optimized prompt with all enhancements */ createOptimizedPrompt( originalPrompt: string, domain: string, clarityScore: number, risks: string[] ): string { let optimized = `**Domain:** ${domain}\n\n`; optimized += `**Original Request:**\n${originalPrompt}\n\n`; // Add domain-specific requirements optimized += `**Requirements Based on Domain:**\n`; switch (domain) { case 'code': optimized += `- Include code summary and complexity notes\n`; optimized += `- Add security considerations\n`; optimized += `- Provide test plan and example I/O\n`; optimized += `- Include error handling\n`; if (risks.includes('security')) { optimized += `- **CRITICAL:** Address security concerns (authentication, validation, data exposure)\n`; } break; case 'UX': optimized += `- Evaluate against usability heuristics\n`; optimized += `- Include accessibility checklist (WCAG 2.1 AA)\n`; optimized += `- Consider mobile responsiveness\n`; optimized += `- Address error and loading states\n`; break; case 'data': optimized += `- Describe dataset shape and structure\n`; optimized += `- Show calculation steps explicitly\n`; optimized += `- Validate data and identify edge cases\n`; optimized += `- Make results reproducible\n`; break; case 'writing': optimized += `- Define audience clearly\n`; optimized += `- Specify tone and style\n`; optimized += `- Include structure outline\n`; optimized += `- Provide strong opening and CTA if applicable\n`; break; case 'finance': optimized += `- State all assumptions explicitly\n`; optimized += `- Show calculation methodology\n`; optimized += `- Identify risk factors\n`; optimized += `- Include sensitivity analysis if relevant\n`; if (risks.includes('compliance')) { optimized += `- **DISCLAIMER:** Not professional financial advice\n`; } break; default: optimized += `- Provide clear, structured output\n`; optimized += `- State assumptions explicitly\n`; optimized += `- Include examples if helpful\n`; } optimized += `\n**Output Format:**\n`; optimized += `- Structured and scannable\n`; optimized += `- Include acceptance criteria\n`; optimized += `- List any assumptions made\n`; if (risks.length > 0) { optimized += `\n**Risk Flags Detected:** ${risks.join(', ')}\n`; optimized += `Please address these concerns in your response.\n`; } return optimized; } /** * Main optimization method */ async optimize(prompt: string, context?: string): Promise<PromptAnalysis> { if (!this.framework) { await this.loadFramework(); } const analysis = this.analyzePrompt(prompt); const domain = analysis.domain!; const clarityScore = analysis.clarity_score!; const risks = analysis.risk_flags!; const questions = this.generateQuestions(prompt, domain, clarityScore); const needsClarification = clarityScore < 0.6 || risks.some(r => ['policy', 'safety'].includes(r)); const optimizedPrompt = this.createOptimizedPrompt(prompt, domain, clarityScore, risks); const header = `[OPTIMIZED] Domain: ${domain} | Clarity: ${(clarityScore * 100).toFixed(0)}% | ` + `Risks: ${risks.length > 0 ? risks.join(', ') : 'none'}`; return { domain, clarity_score: clarityScore, risk_flags: risks, questions, optimized_prompt: optimizedPrompt, optimization_header: header, assumptions: [], needs_clarification: needsClarification, }; } } // Create the MCP server const server = new Server( { name: 'prompt-optimizer', version: '1.0.0', }, { capabilities: { tools: {}, }, } ); const optimizer = new PromptOptimizer(); // Define the optimize_prompt tool const OPTIMIZE_TOOL: Tool = { name: 'optimize_prompt', description: 'Analyze and optimize a user prompt using the OTA (Optimize-Then-Answer) Framework. ' + 'Returns clarity score, domain classification, risk flags, targeted questions (if needed), ' + 'and an enhanced prompt ready for AI processing.', inputSchema: { type: 'object', properties: { prompt: { type: 'string', description: 'The user prompt to optimize', }, context: { type: 'string', description: 'Optional additional context about the request', }, }, required: ['prompt'], }, }; // List available tools server.setRequestHandler(ListToolsRequestSchema, async () => { return { tools: [OPTIMIZE_TOOL], }; }); // Handle tool calls server.setRequestHandler(CallToolRequestSchema, async (request) => { if (request.params.name === 'optimize_prompt') { const { prompt, context } = request.params.arguments as { prompt: string; context?: string }; try { const result = await optimizer.optimize(prompt, context); let response = `${result.optimization_header}\n\n`; if (result.needs_clarification && result.questions.length > 0) { response += `**⚠️ Clarification Needed** (Clarity: ${(result.clarity_score * 100).toFixed(0)}%)\n\n`; response += `**Please answer these questions before I proceed:**\n\n`; result.questions.forEach((q, i) => { response += `${i + 1}. ${q}\n`; }); response += `\n---\n\n`; } else { response += `**✓ Ready to Process** (Clarity: ${(result.clarity_score * 100).toFixed(0)}%)\n\n`; } response += `**Domain:** ${result.domain}\n`; response += `**Risk Flags:** ${result.risk_flags.length > 0 ? result.risk_flags.join(', ') : 'None'}\n\n`; if (!result.needs_clarification) { response += `**Optimized Prompt:**\n\`\`\`\n${result.optimized_prompt}\n\`\`\`\n\n`; response += `Use this enhanced prompt for the AI request to ensure comprehensive, domain-appropriate output.`; } return { content: [ { type: 'text', text: response, }, ], }; } catch (error) { return { content: [ { type: 'text', text: `Error optimizing prompt: ${error}`, }, ], isError: true, }; } } throw new Error(`Unknown tool: ${request.params.name}`); }); // Start the server async function main() { const transport = new StdioServerTransport(); await server.connect(transport); console.error('Prompt Optimizer MCP Server running on stdio'); } main().catch((error) => { console.error('Fatal error:', error); process.exit(1); });