ai-cost-analyzer 1.0.0

package/dist/types.d.ts

@@ -0,0 +1,144 @@
+/**
+ * Shared types for the AI Cost Analyzer MCP server.
+ */
+export interface ModelPricing {
+    /** Human-readable model name */
+    name: string;
+    /** Provider (e.g. "Anthropic", "OpenAI", "Google") */
+    provider: string;
+    /** Cost per 1M input tokens in USD */
+    inputPerMillion: number;
+    /** Cost per 1M output tokens in USD */
+    outputPerMillion: number;
+    /** Cost per 1M cached input tokens in USD (if supported) */
+    cachedInputPerMillion?: number;
+    /** Max context window in tokens */
+    contextWindow: number;
+    /** Max output tokens */
+    maxOutput: number;
+    /** Relative quality tier: 1 = frontier, 2 = strong, 3 = fast/cheap */
+    qualityTier: 1 | 2 | 3;
+}
+export type ModelId = "claude-opus" | "claude-sonnet" | "claude-haiku" | "gpt-4o" | "gpt-4o-mini" | "gemini-2.0-flash";
+export interface UsageRecord {
+    /** Model identifier */
+    model: string;
+    /** Number of input tokens */
+    input_tokens: number;
+    /** Number of output tokens */
+    output_tokens: number;
+    /** Number of cached input tokens (subset of input_tokens) */
+    cached_tokens?: number;
+    /** ISO-8601 timestamp */
+    timestamp: string;
+    /** Optional: number of tool definitions sent in the request */
+    tool_definitions?: number;
+    /** Optional: system prompt token count */
+    system_prompt_tokens?: number;
+}
+export interface ModelCostBreakdown {
+    model: string;
+    requestCount: number;
+    totalInputTokens: number;
+    totalOutputTokens: number;
+    totalCachedTokens: number;
+    inputCost: number;
+    outputCost: number;
+    cacheSavings: number;
+    totalCost: number;
+}
+export interface TrendDataPoint {
+    period: string;
+    requestCount: number;
+    totalCost: number;
+    avgCostPerRequest: number;
+}
+export interface UsageAnalysis {
+    summary: {
+        totalRequests: number;
+        totalCost: number;
+        avgCostPerRequest: number;
+        totalInputTokens: number;
+        totalOutputTokens: number;
+        totalCachedTokens: number;
+    };
+    costByModel: ModelCostBreakdown[];
+    wasteEstimation: {
+        unusedToolDefinitionTokens: number;
+        oversizedSystemPromptTokens: number;
+        estimatedWasteCost: number;
+        wastePercentage: number;
+    };
+    trends: {
+        daily: TrendDataPoint[];
+        weekly: TrendDataPoint[];
+    };
+}
+export interface SavingsEstimate {
+    currentMonthlyCost: number;
+    optimizedMonthlyCost: number;
+    totalMonthlySavings: number;
+    savingsPercentage: number;
+    breakdown: {
+        promptCaching: {
+            description: string;
+            currentCost: number;
+            optimizedCost: number;
+            monthlySavings: number;
+        };
+        modelRouting: {
+            description: string;
+            currentCost: number;
+            optimizedCost: number;
+            monthlySavings: number;
+        };
+        contextPruning: {
+            description: string;
+            currentCost: number;
+            optimizedCost: number;
+            monthlySavings: number;
+        };
+    };
+    recommendations: string[];
+}
+export interface OptimizationResult {
+    original: {
+        systemPrompt: string;
+        systemPromptTokens: number;
+        toolDefinitions: string[];
+        toolDefinitionTokens: number;
+        totalTokens: number;
+    };
+    optimized: {
+        systemPrompt: string;
+        systemPromptTokens: number;
+        toolDefinitions: string[];
+        toolDefinitionTokens: number;
+        totalTokens: number;
+    };
+    savings: {
+        tokensReduced: number;
+        percentageReduction: number;
+        estimatedMonthlySavings: number;
+    };
+    recommendations: string[];
+}
+export interface ModelComparisonEntry {
+    model: string;
+    provider: string;
+    qualityTier: string;
+    estimatedInputCost: number;
+    estimatedOutputCost: number;
+    estimatedTotalCost: number;
+    costPer1000Requests: number;
+    recommendation: string;
+    bestFor: string[];
+}
+export interface ModelComparison {
+    taskDescription: string;
+    estimatedInputTokens: number;
+    estimatedOutputTokens: number;
+    models: ModelComparisonEntry[];
+    recommendation: string;
+}
+//# sourceMappingURL=types.d.ts.map

package/dist/types.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA;;GAEG;AAMH,MAAM,WAAW,YAAY;IAC3B,gCAAgC;IAChC,IAAI,EAAE,MAAM,CAAC;IACb,sDAAsD;IACtD,QAAQ,EAAE,MAAM,CAAC;IACjB,sCAAsC;IACtC,eAAe,EAAE,MAAM,CAAC;IACxB,uCAAuC;IACvC,gBAAgB,EAAE,MAAM,CAAC;IACzB,4DAA4D;IAC5D,qBAAqB,CAAC,EAAE,MAAM,CAAC;IAC/B,mCAAmC;IACnC,aAAa,EAAE,MAAM,CAAC;IACtB,wBAAwB;IACxB,SAAS,EAAE,MAAM,CAAC;IAClB,sEAAsE;IACtE,WAAW,EAAE,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;CACxB;AAED,MAAM,MAAM,OAAO,GACf,aAAa,GACb,eAAe,GACf,cAAc,GACd,QAAQ,GACR,aAAa,GACb,kBAAkB,CAAC;AAMvB,MAAM,WAAW,WAAW;IAC1B,uBAAuB;IACvB,KAAK,EAAE,MAAM,CAAC;IACd,6BAA6B;IAC7B,YAAY,EAAE,MAAM,CAAC;IACrB,8BAA8B;IAC9B,aAAa,EAAE,MAAM,CAAC;IACtB,6DAA6D;IAC7D,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,yBAAyB;IACzB,SAAS,EAAE,MAAM,CAAC;IAClB,+DAA+D;IAC/D,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,0CAA0C;IAC1C,oBAAoB,CAAC,EAAE,MAAM,CAAC;CAC/B;AAMD,MAAM,WAAW,kBAAkB;IACjC,KAAK,EAAE,MAAM,CAAC;IACd,YAAY,EAAE,MAAM,CAAC;IACrB,gBAAgB,EAAE,MAAM,CAAC;IACzB,iBAAiB,EAAE,MAAM,CAAC;IAC1B,iBAAiB,EAAE,MAAM,CAAC;IAC1B,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,YAAY,EAAE,MAAM,CAAC;IACrB,SAAS,EAAE,MAAM,CAAC;CACnB;AAED,MAAM,WAAW,cAAc;IAC7B,MAAM,EAAE,MAAM,CAAC;IACf,YAAY,EAAE,MAAM,CAAC;IACrB,SAAS,EAAE,MAAM,CAAC;IAClB,iBAAiB,EAAE,MAAM,CAAC;CAC3B;AAED,MAAM,WAAW,aAAa;IAC5B,OAAO,EAAE;QACP,aAAa,EAAE,MAAM,CAAC;QACtB,SAAS,EAAE,MAAM,CAAC;QAClB,iBAAiB,EAAE,MAAM,CAAC;QAC1B,gBAAgB,EAAE,MAAM,CAAC;QACzB,iBAAiB,EAAE,MAAM,CAAC;QAC1B,iBAAiB,EAAE,MAAM,CAAC;KAC3B,CAAC;IACF,WAAW,EAAE,kBAAkB,EAAE,CAAC;IAClC,eAAe,EAAE;QACf,0BAA0B,EAAE,MAAM,CAAC;QACnC,2BAA2B,EAAE,MAAM,CAAC;QACpC,kBAAkB,EAAE,MAAM,CAAC;QAC3B,eAAe,EAAE,MAAM,CAAC;KACzB,CAAC;IACF,MAAM,EAAE;QACN,KAAK,EAAE,cAAc,EAAE,CAAC;QACxB,MAAM,EAAE,cAAc,EAAE,CAAC;KAC1B,CAAC;CACH;AAMD,MAAM,WAAW,eAAe;IAC9B,kBAAkB,EAAE,MAAM,CAAC;IAC3B,oBAAoB,EAAE,MAAM,CAAC;IAC7B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,iBAAiB,EAAE,MAAM,CAAC;IAC1B,SAAS,EAAE;QACT,aAAa,EAAE;YACb,WAAW,EAAE,MAAM,CAAC;YACpB,WAAW,EAAE,MAAM,CAAC;YACpB,aAAa,EAAE,MAAM,CAAC;YACtB,cAAc,EAAE,MAAM,CAAC;SACxB,CAAC;QACF,YAAY,EAAE;YACZ,WAAW,EAAE,MAAM,CAAC;YACpB,WAAW,EAAE,MAAM,CAAC;YACpB,aAAa,EAAE,MAAM,CAAC;YACtB,cAAc,EAAE,MAAM,CAAC;SACxB,CAAC;QACF,cAAc,EAAE;YACd,WAAW,EAAE,MAAM,CAAC;YACpB,WAAW,EAAE,MAAM,CAAC;YACpB,aAAa,EAAE,MAAM,CAAC;YACtB,cAAc,EAAE,MAAM,CAAC;SACxB,CAAC;KACH,CAAC;IACF,eAAe,EAAE,MAAM,EAAE,CAAC;CAC3B;AAMD,MAAM,WAAW,kBAAkB;IACjC,QAAQ,EAAE;QACR,YAAY,EAAE,MAAM,CAAC;QACrB,kBAAkB,EAAE,MAAM,CAAC;QAC3B,eAAe,EAAE,MAAM,EAAE,CAAC;QAC1B,oBAAoB,EAAE,MAAM,CAAC;QAC7B,WAAW,EAAE,MAAM,CAAC;KACrB,CAAC;IACF,SAAS,EAAE;QACT,YAAY,EAAE,MAAM,CAAC;QACrB,kBAAkB,EAAE,MAAM,CAAC;QAC3B,eAAe,EAAE,MAAM,EAAE,CAAC;QAC1B,oBAAoB,EAAE,MAAM,CAAC;QAC7B,WAAW,EAAE,MAAM,CAAC;KACrB,CAAC;IACF,OAAO,EAAE;QACP,aAAa,EAAE,MAAM,CAAC;QACtB,mBAAmB,EAAE,MAAM,CAAC;QAC5B,uBAAuB,EAAE,MAAM,CAAC;KACjC,CAAC;IACF,eAAe,EAAE,MAAM,EAAE,CAAC;CAC3B;AAMD,MAAM,WAAW,oBAAoB;IACnC,KAAK,EAAE,MAAM,CAAC;IACd,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,kBAAkB,EAAE,MAAM,CAAC;IAC3B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,kBAAkB,EAAE,MAAM,CAAC;IAC3B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,cAAc,EAAE,MAAM,CAAC;IACvB,OAAO,EAAE,MAAM,EAAE,CAAC;CACnB;AAED,MAAM,WAAW,eAAe;IAC9B,eAAe,EAAE,MAAM,CAAC;IACxB,oBAAoB,EAAE,MAAM,CAAC;IAC7B,qBAAqB,EAAE,MAAM,CAAC;IAC9B,MAAM,EAAE,oBAAoB,EAAE,CAAC;IAC/B,cAAc,EAAE,MAAM,CAAC;CACxB"}

package/dist/types.js

@@ -0,0 +1,5 @@
+/**
+ * Shared types for the AI Cost Analyzer MCP server.
+ */
+export {};
+//# sourceMappingURL=types.js.map

package/dist/types.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA;;GAEG"}

package/package.json

@@ -0,0 +1,41 @@
+{
+  "name": "ai-cost-analyzer",
+  "version": "1.0.0",
+  "description": "MCP server that helps developers understand and reduce their AI API costs through usage analysis, savings estimation, prompt optimization, and model comparison.",
+  "type": "module",
+  "main": "dist/index.js",
+  "bin": {
+    "ai-cost-analyzer": "dist/index.js"
+  },
+  "scripts": {
+    "build": "tsc",
+    "start": "node dist/index.js",
+    "start:sse": "node dist/index.js --sse",
+    "dev": "tsc --watch",
+    "prepublishOnly": "npm run build"
+  },
+  "keywords": [
+    "mcp",
+    "ai",
+    "cost",
+    "analyzer",
+    "llm",
+    "tokens",
+    "optimization",
+    "claude",
+    "openai",
+    "gemini"
+  ],
+  "license": "MIT",
+  "dependencies": {
+    "@modelcontextprotocol/sdk": "^1.12.0",
+    "zod": "^3.25.0"
+  },
+  "devDependencies": {
+    "@types/node": "^22.0.0",
+    "typescript": "^5.7.0"
+  },
+  "engines": {
+    "node": ">=18.0.0"
+  }
+}

package/src/index.ts

@@ -0,0 +1,291 @@
+#!/usr/bin/env node
+
+/**
+ * AI Cost Analyzer -- MCP Server
+ *
+ * Helps developers understand and reduce their AI API costs through
+ * usage analysis, savings estimation, prompt optimization, and model comparison.
+ */
+
+import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
+import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
+import { SSEServerTransport } from "@modelcontextprotocol/sdk/server/sse.js";
+import { z } from "zod";
+import http from "node:http";
+
+import { handleAnalyzeUsage } from "./tools/analyze.js";
+import { handleEstimateSavings } from "./tools/savings.js";
+import { handleOptimizePrompt } from "./tools/optimize.js";
+import { handleCompareModels } from "./tools/compare.js";
+import { handleGetPricing } from "./tools/pricing.js";
+
+// ---------------------------------------------------------------------------
+// Zod schemas for tool inputs
+// ---------------------------------------------------------------------------
+
+const UsageRecordSchema = z.object({
+  model: z.string().describe("Model identifier (e.g. 'claude-sonnet', 'gpt-4o')"),
+  input_tokens: z.number().int().min(0).describe("Number of input tokens"),
+  output_tokens: z.number().int().min(0).describe("Number of output tokens"),
+  cached_tokens: z.number().int().min(0).optional().describe("Number of cached input tokens"),
+  timestamp: z.string().describe("ISO-8601 timestamp of the request"),
+  tool_definitions: z.number().int().min(0).optional().describe("Number of tool definitions sent"),
+  system_prompt_tokens: z.number().int().min(0).optional().describe("Token count of the system prompt"),
+});
+
+// ---------------------------------------------------------------------------
+// Server setup
+// ---------------------------------------------------------------------------
+
+const server = new McpServer({
+  name: "ai-cost-analyzer",
+  version: "1.0.0",
+});
+
+// ---------------------------------------------------------------------------
+// Tool: analyze_usage
+// ---------------------------------------------------------------------------
+
+server.tool(
+  "analyze_usage",
+  "Analyze AI API usage data and return cost breakdowns by model, average cost per request, token waste estimation, and daily/weekly trends.",
+  {
+    usage_data: z.array(UsageRecordSchema).describe(
+      "Array of API usage records with model, tokens, and timestamps",
+    ),
+  },
+  async ({ usage_data }) => {
+    try {
+      const result = handleAnalyzeUsage(usage_data);
+      return {
+        content: [{ type: "text" as const, text: result }],
+      };
+    } catch (err) {
+      return {
+        content: [{
+          type: "text" as const,
+          text: `Error analyzing usage: ${err instanceof Error ? err.message : String(err)}`,
+        }],
+        isError: true,
+      };
+    }
+  },
+);
+
+// ---------------------------------------------------------------------------
+// Tool: estimate_savings
+// ---------------------------------------------------------------------------
+
+server.tool(
+  "estimate_savings",
+  "Estimate potential monthly savings from prompt caching, model routing, and context pruning based on current usage patterns.",
+  {
+    usage_data: z.array(UsageRecordSchema).describe(
+      "Array of API usage records representing current usage patterns",
+    ),
+    monthly_multiplier: z.number().min(1).optional().describe(
+      "If the usage_data is a sample, multiply costs by this factor to project monthly totals (default: 1)",
+    ),
+  },
+  async ({ usage_data, monthly_multiplier }) => {
+    try {
+      const result = handleEstimateSavings(usage_data, monthly_multiplier);
+      return {
+        content: [{ type: "text" as const, text: result }],
+      };
+    } catch (err) {
+      return {
+        content: [{
+          type: "text" as const,
+          text: `Error estimating savings: ${err instanceof Error ? err.message : String(err)}`,
+        }],
+        isError: true,
+      };
+    }
+  },
+);
+
+// ---------------------------------------------------------------------------
+// Tool: optimize_prompt
+// ---------------------------------------------------------------------------
+
+server.tool(
+  "optimize_prompt",
+  "Analyze a system prompt and tool definitions, return an optimized version with redundant content removed, token counts before/after, and cost reduction estimates.",
+  {
+    system_prompt: z.string().describe("The system prompt text to optimize"),
+    tool_definitions: z.array(z.string()).describe(
+      "Array of tool definition strings (JSON or plain text)",
+    ),
+    requests_per_month: z.number().int().min(1).optional().describe(
+      "Estimated requests per month for cost projection (default: 10,000)",
+    ),
+    model: z.string().optional().describe(
+      "Model name for pricing calculation (default: claude-sonnet)",
+    ),
+  },
+  async ({ system_prompt, tool_definitions, requests_per_month, model }) => {
+    try {
+      const result = handleOptimizePrompt(system_prompt, tool_definitions, requests_per_month, model);
+      return {
+        content: [{ type: "text" as const, text: result }],
+      };
+    } catch (err) {
+      return {
+        content: [{
+          type: "text" as const,
+          text: `Error optimizing prompt: ${err instanceof Error ? err.message : String(err)}`,
+        }],
+        isError: true,
+      };
+    }
+  },
+);
+
+// ---------------------------------------------------------------------------
+// Tool: compare_models
+// ---------------------------------------------------------------------------
+
+server.tool(
+  "compare_models",
+  "Compare cost and quality tradeoffs across Claude, GPT-4o, GPT-4o-mini, and Gemini models for a given task. Returns cost per request, monthly projections, and recommendations.",
+  {
+    task_description: z.string().describe(
+      "Description of the task to compare models for",
+    ),
+    estimated_input_tokens: z.number().int().min(1).optional().describe(
+      "Override estimated input tokens per request",
+    ),
+    estimated_output_tokens: z.number().int().min(1).optional().describe(
+      "Override estimated output tokens per request",
+    ),
+  },
+  async ({ task_description, estimated_input_tokens, estimated_output_tokens }) => {
+    try {
+      const result = handleCompareModels(
+        task_description,
+        estimated_input_tokens,
+        estimated_output_tokens,
+      );
+      return {
+        content: [{ type: "text" as const, text: result }],
+      };
+    } catch (err) {
+      return {
+        content: [{
+          type: "text" as const,
+          text: `Error comparing models: ${err instanceof Error ? err.message : String(err)}`,
+        }],
+        isError: true,
+      };
+    }
+  },
+);
+
+// ---------------------------------------------------------------------------
+// Tool: get_pricing
+// ---------------------------------------------------------------------------
+
+server.tool(
+  "get_pricing",
+  "Return current pricing for all major LLM APIs (Claude, OpenAI, Gemini) with input, output, and cached token rates per million tokens.",
+  {},
+  async () => {
+    try {
+      const result = handleGetPricing();
+      return {
+        content: [{ type: "text" as const, text: result }],
+      };
+    } catch (err) {
+      return {
+        content: [{
+          type: "text" as const,
+          text: `Error fetching pricing: ${err instanceof Error ? err.message : String(err)}`,
+        }],
+        isError: true,
+      };
+    }
+  },
+);
+
+// ---------------------------------------------------------------------------
+// Transport: stdio or SSE
+// ---------------------------------------------------------------------------
+
+async function main(): Promise<void> {
+  const useSSE = process.argv.includes("--sse");
+
+  if (useSSE) {
+    const port = parseInt(process.env.PORT ?? "3000", 10);
+
+    // Track active transports for cleanup
+    const transports = new Map<string, SSEServerTransport>();
+
+    const httpServer = http.createServer(async (req, res) => {
+      // CORS headers
+      res.setHeader("Access-Control-Allow-Origin", "*");
+      res.setHeader("Access-Control-Allow-Methods", "GET, POST, OPTIONS");
+      res.setHeader("Access-Control-Allow-Headers", "Content-Type");
+
+      if (req.method === "OPTIONS") {
+        res.writeHead(204);
+        res.end();
+        return;
+      }
+
+      const url = new URL(req.url ?? "/", `http://localhost:${port}`);
+
+      if (url.pathname === "/sse" && req.method === "GET") {
+        // Create a new SSE transport for this connection
+        const transport = new SSEServerTransport("/messages", res);
+        const sessionId = transport.sessionId;
+        transports.set(sessionId, transport);
+
+        res.on("close", () => {
+          transports.delete(sessionId);
+        });
+
+        await server.connect(transport);
+        return;
+      }
+
+      if (url.pathname === "/messages" && req.method === "POST") {
+        const sessionId = url.searchParams.get("sessionId");
+        if (!sessionId || !transports.has(sessionId)) {
+          res.writeHead(400, { "Content-Type": "application/json" });
+          res.end(JSON.stringify({ error: "Invalid or missing sessionId" }));
+          return;
+        }
+        const transport = transports.get(sessionId)!;
+        await transport.handlePostMessage(req, res);
+        return;
+      }
+
+      // Health check
+      if (url.pathname === "/health") {
+        res.writeHead(200, { "Content-Type": "application/json" });
+        res.end(JSON.stringify({ status: "ok", server: "ai-cost-analyzer", version: "1.0.0" }));
+        return;
+      }
+
+      res.writeHead(404, { "Content-Type": "application/json" });
+      res.end(JSON.stringify({ error: "Not found" }));
+    });
+
+    httpServer.listen(port, () => {
+      console.error(`AI Cost Analyzer MCP server (SSE) listening on port ${port}`);
+      console.error(`  SSE endpoint: http://localhost:${port}/sse`);
+      console.error(`  Messages endpoint: http://localhost:${port}/messages`);
+      console.error(`  Health check: http://localhost:${port}/health`);
+    });
+  } else {
+    const transport = new StdioServerTransport();
+    await server.connect(transport);
+    console.error("AI Cost Analyzer MCP server running on stdio");
+  }
+}
+
+main().catch((err) => {
+  console.error("Fatal error:", err);
+  process.exit(1);
+});