MCP AI Bug Helper

#!/usr/bin/env node import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'; import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'; import { z } from 'zod'; import pkg from '../package.json' with { type: 'json' }; import { config, assertConfig } from './config.js'; import { OpenRouterClient } from './openrouterClient.js'; import { CodingAdvisorCoordinator } from './codingAdvisors.js'; import type { ToolInput, AdvisorBatchResult } from './types.js'; import { RateLimitError, OpenRouterError } from './errors.js'; assertConfig(); const client = new OpenRouterClient(config.apiKey, { appName: config.appName, referer: config.referer, }); const coordinator = new CodingAdvisorCoordinator(client); const server = new McpServer( { name: 'coding-advisor-mcp', version: typeof pkg.version === 'string' ? pkg.version : '0.0.0', }, { capabilities: { tools: {}, }, instructions: 'Use the ask-coding-advisors tool whenever you want three independent coding-specialist LLM takes via OpenRouter (free first, paid fallback after rate limits). Provide the bug or question plus any context.', } ); const inputSchema = z.object({ question: z .string() .min(4, 'Bitte fasse dein Problem in mindestens vier Zeichen zusammen.') .max(4000, 'Question too long (4k char cap).'), context: z.string().max(12000).optional(), }); server.registerTool( 'ask-coding-advisors', { title: 'Ask Coding Advisors', description: 'Queries three OpenRouter coding specialists (free tier) and falls back to premium models when rate limited.', inputSchema, }, async (args) => { const input: ToolInput = typeof args.context === 'string' && args.context.length > 0 ? { question: args.question, context: args.context } : { question: args.question }; try { const batch = await coordinator.advise(input); const formatted = formatBatch(batch); return { content: [ { type: 'text', text: formatted, }, ], }; } catch (error) { const prefix = 'Coding advisor tool failed'; const message = error instanceof Error ? `${prefix}: ${error.message}` : `${prefix}: ${String(error)}`; const isRetryable = error instanceof RateLimitError || error instanceof OpenRouterError; return { isError: isRetryable, content: [ { type: 'text', text: message, }, ], }; } } ); const transport = new StdioServerTransport(); server .connect(transport) .catch((error) => { console.error('Failed to start MCP server', error); process.exit(1); }); function formatBatch(batch: AdvisorBatchResult): string { const header = batch.fallbackTriggered ? '⚠️ OpenRouter rate limit hit — switched to paid fallbacks for the remaining slots.\n' : ''; const body = batch.answers .map((answer, index) => { const tier = answer.isFree ? 'free tier' : answer.usedFallback ? 'paid fallback' : 'paid'; const latency = formatLatency(answer.latencyMs); const usageBits = formatUsage(answer.usage); return [ `Advisor ${index + 1}: ${answer.modelLabel} (${tier})`, `Focus: ${answer.focus}`, `Latency: ${latency}${usageBits ? ` | Usage: ${usageBits}` : ''}`, '', answer.responseText, ].join('\n'); }) .join('\n\n---\n\n'); return header + body; } function formatLatency(ms: number): string { if (!Number.isFinite(ms)) { return 'n/a'; } if (ms < 1000) { return `${ms.toFixed(0)} ms`; } return `${(ms / 1000).toFixed(2)} s`; } function formatUsage(usage?: { promptTokens?: number; completionTokens?: number; totalTokens?: number }) { if (!usage) return ''; const parts: string[] = []; if (Number.isFinite(usage.promptTokens)) { parts.push(`prompt ${usage.promptTokens}`); } if (Number.isFinite(usage.completionTokens)) { parts.push(`completion ${usage.completionTokens}`); } if (Number.isFinite(usage.totalTokens)) { parts.push(`total ${usage.totalTokens}`); } return parts.join(', '); }