AI Agent Template MCP Server

Instructions

Implementation Reference

• src/tools/performance/track-performance.ts:48-115 (handler)

  The core handler function that processes input metrics, calculates scores, generates reports with recommendations, trends, and achievements, and updates agent memory/context files.
  export async function trackAgentPerformance( metrics: PerformanceMetrics ): Promise<PerformanceReport> { const report: PerformanceReport = { success: false, summary: { overallScore: 0, efficiency: '', quality: '', security: '', }, recommendations: [], trends: { tokenUsage: 'stable', quality: 'stable', speed: 'stable', }, achievements: [], }; try { // Calculate overall score const scores = { validation: metrics.metrics.validationScore, security: metrics.metrics.securityScore, coverage: metrics.metrics.testCoverage, efficiency: calculateEfficiencyScore(metrics), errorFree: calculateErrorScore(metrics), }; const overallScore = Math.round( (scores.validation * 0.25 + scores.security * 0.25 + scores.coverage * 0.2 + scores.efficiency * 0.15 + scores.errorFree * 0.15) ); // Update report summary report.summary.overallScore = overallScore; report.summary.efficiency = getEfficiencyRating(metrics); report.summary.quality = getQualityRating(scores.validation, scores.coverage); report.summary.security = getSecurityRating(scores.security); // Update agent memory with performance data updateAgentMemory(metrics, overallScore); // Analyze trends report.trends = analyzeTrends(metrics); // Generate recommendations report.recommendations = generateRecommendations(metrics, scores); // Check for achievements report.achievements = checkAchievements(metrics, overallScore); // Update context with current metrics updateAgentContext(metrics); report.success = true; } catch (error) { report.success = false; report.recommendations = [`Error tracking performance: ${error}`]; } return report; }
• src/tools/tool-definitions.ts:182-237 (schema)

  MCP tool definition with input schema specifying the structure for performance metrics and optional improvements.
  { name: 'track_agent_performance', description: 'Track and analyze AI agent performance metrics', inputSchema: { type: 'object', properties: { featureName: { type: 'string', description: 'Name of the feature being tracked', }, timestamp: { type: 'string', description: 'ISO timestamp of the feature completion', }, metrics: { type: 'object', properties: { tokensUsed: { type: 'number' }, timeElapsed: { type: 'number' }, validationScore: { type: 'number' }, securityScore: { type: 'number' }, testCoverage: { type: 'number' }, hallucinations: { type: 'object', properties: { detected: { type: 'number' }, prevented: { type: 'number' }, examples: { type: 'array', items: { type: 'string' }, }, }, }, errors: { type: 'object', properties: { syntax: { type: 'number' }, runtime: { type: 'number' }, type: { type: 'number' }, }, }, }, required: ['tokensUsed', 'timeElapsed', 'validationScore', 'securityScore', 'testCoverage'], }, improvements: { type: 'object', properties: { tokenReduction: { type: 'number' }, timeReduction: { type: 'number' }, qualityIncrease: { type: 'number' }, }, }, }, required: ['featureName', 'timestamp', 'metrics'], }, },
• src/tools/index.ts:185-230 (registration)

  Tool registration in the MCP server handler: parses arguments with Zod schema matching the tool definition and calls the trackAgentPerformance handler.
  case 'track_agent_performance': { const params = z.object({ featureName: z.string(), timestamp: z.string(), metrics: z.object({ tokensUsed: z.number(), timeElapsed: z.number(), validationScore: z.number(), securityScore: z.number(), testCoverage: z.number(), hallucinations: z.object({ detected: z.number(), prevented: z.number(), examples: z.array(z.string()), }).optional().default({ detected: 0, prevented: 0, examples: [], }), errors: z.object({ syntax: z.number(), runtime: z.number(), type: z.number(), }).optional().default({ syntax: 0, runtime: 0, type: 0, }), }), improvements: z.object({ tokenReduction: z.number().optional(), timeReduction: z.number().optional(), qualityIncrease: z.number().optional(), }).optional(), }).parse(args); const result = await trackAgentPerformance(params); return { content: [ { type: 'text', text: JSON.stringify(result, null, 2), }, ], }; }
• src/tools/performance/track-performance.ts:4-29 (schema)

  TypeScript interface defining the input structure for PerformanceMetrics used by the handler.
  interface PerformanceMetrics { featureName: string; timestamp: string; metrics: { tokensUsed: number; timeElapsed: number; // in seconds validationScore: number; // 0-100 securityScore: number; // 0-100 testCoverage: number; // percentage hallucinations: { detected: number; prevented: number; examples: string[]; }; errors: { syntax: number; runtime: number; type: number; }; }; improvements?: { tokenReduction?: number; timeReduction?: number; qualityIncrease?: number; }; }
• src/tools/performance/track-performance.ts:31-46 (schema)

  TypeScript interface defining the output structure for PerformanceReport returned by the handler.
  interface PerformanceReport { success: boolean; summary: { overallScore: number; efficiency: string; quality: string; security: string; }; recommendations: string[]; trends: { tokenUsage: 'improving' | 'stable' | 'degrading'; quality: 'improving' | 'stable' | 'degrading'; speed: 'improving' | 'stable' | 'degrading'; }; achievements: string[]; }