Langfuse MCP Server

#!/usr/bin/env node // Parse command line arguments for mode selection function parseCommandLineMode(): 'readonly' | 'readwrite' | null { const args = process.argv.slice(2); // Check for explicit mode flag: --mode=readonly or --mode=readwrite const modeFlag = args.find(arg => arg.startsWith('--mode=')); if (modeFlag) { const mode = modeFlag.split('=')[1]?.toLowerCase(); if (mode === 'readonly' || mode === 'readwrite') { return mode; } } // Check for readwrite flag: --readwrite if (args.includes('--readwrite') || args.includes('--rw')) { return 'readwrite'; } // Check for readonly flag: --readonly or --ro (explicit readonly) if (args.includes('--readonly') || args.includes('--ro')) { return 'readonly'; } // No explicit mode specified via CLI return null; } // Set mode based on CLI args, fallback to environment variables, default to readonly const cliMode = parseCommandLineMode(); if (cliMode) { process.env.LANGFUSE_MCP_MODE = cliMode; } else if (!process.env.LANGFUSE_MCP_MODE) { // Safe default: readonly mode if no explicit configuration process.env.LANGFUSE_MCP_MODE = 'readonly'; } import { Server } from '@modelcontextprotocol/sdk/server/index.js'; import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'; import { CallToolRequestSchema, ListToolsRequestSchema, } from '@modelcontextprotocol/sdk/types.js'; import { getProjectConfig, getServerModeConfig, getCurrentServerMode } from './config.js'; import { LangfuseAnalyticsClient } from './langfuse-client.js'; import { ServerMode, ServerModeConfig } from './types.js'; import { isToolAllowed, isWriteTool, assertWriteEnabled, validateConfirmation } from './mode-config.js'; import { AuditLogger } from './audit-logger.js'; // Import tool handlers import { listProjects, listProjectsSchema } from './tools/list-projects.js'; import { projectOverview, projectOverviewSchema } from './tools/project-overview.js'; import { usageByModel, usageByModelSchema } from './tools/usage-by-model.js'; import { usageByService, usageByServiceSchema } from './tools/usage-by-service.js'; import { topExpensiveTraces, topExpensiveTracesSchema } from './tools/top-expensive-traces.js'; import { getTraceDetail, getTraceDetailSchema } from './tools/get-trace-detail.js'; // New requested tools import { getMetrics, getMetricsSchema } from './tools/get-metrics.js'; import { getTraces, getTracesSchema } from './tools/get-traces.js'; import { getObservations, getObservationsSchema } from './tools/get-observations.js'; import { getCostAnalysis, getCostAnalysisSchema } from './tools/get-cost-analysis.js'; import { getDailyMetrics, getDailyMetricsSchema } from './tools/get-daily-metrics.js'; import { getProjects, getProjectsSchema } from './tools/get-projects.js'; // Additional API tools import { getObservationDetail, getObservationDetailSchema } from './tools/get-observation-detail.js'; import { getHealthStatus, getHealthStatusSchema } from './tools/get-health-status.js'; import { listModels, listModelsSchema } from './tools/list-models.js'; import { getModelDetail, getModelDetailSchema } from './tools/get-model-detail.js'; import { listPrompts, listPromptsSchema } from './tools/list-prompts.js'; import { getPromptDetail, getPromptDetailSchema } from './tools/get-prompt-detail.js'; // Dataset management tools import { createDataset, createDatasetSchema } from './tools/create-dataset.js'; import { listDatasets, listDatasetsSchema } from './tools/list-datasets.js'; import { getDataset, getDatasetSchema } from './tools/get-dataset.js'; import { createDatasetItem, createDatasetItemSchema } from './tools/create-dataset-item.js'; import { listDatasetItems, listDatasetItemsSchema } from './tools/list-dataset-items.js'; import { getDatasetItem, getDatasetItemSchema } from './tools/get-dataset-item.js'; import { deleteDatasetItem, deleteDatasetItemSchema } from './tools/delete-dataset-item.js'; // Comment management tools import { createComment, createCommentSchema } from './tools/create-comment.js'; import { listComments, listCommentsSchema } from './tools/list-comments.js'; import { getComment, getCommentSchema } from './tools/get-comment.js'; class LangfuseAnalyticsServer { private server: Server; private client: LangfuseAnalyticsClient; private mode: ServerMode; private modeConfig: ServerModeConfig; private auditLogger: AuditLogger; constructor() { // Initialize mode configuration this.mode = getCurrentServerMode(); this.modeConfig = getServerModeConfig(); this.auditLogger = AuditLogger.getInstance(); this.server = new Server( { name: `langfuse-analytics-${this.mode}`, version: '1.4.0', }, { capabilities: { tools: {}, }, } ); // Initialize Langfuse client const config = getProjectConfig(); this.client = new LangfuseAnalyticsClient(config); // Log mode initialization this.logModeStartup(); this.setupHandlers(); this.setupErrorHandling(); } private logModeStartup(): void { console.error(`🔒 Langfuse MCP Server v1.4.0 - Running in ${this.mode.toUpperCase()} mode`); if (this.mode === 'readonly') { console.error(' ✅ Only read operations allowed. This is the safe default mode.'); console.error(' 💡 Set LANGFUSE_MCP_MODE=readwrite to enable write operations.'); } else { console.error(' ⚠️ Write operations enabled - this can modify your Langfuse data.'); console.error(' 📝 All write operations will be logged for audit purposes.'); } // Log to audit logger this.auditLogger.logModeInitialization(this.mode, this.modeConfig.allowedTools.size); } private setupErrorHandling(): void { this.server.onerror = (error) => { console.error('[MCP Error]', error); }; process.on('SIGINT', async () => { await this.client.shutdown(); process.exit(0); }); } private setupHandlers(): void { // List tools handler - filter tools based on current mode this.server.setRequestHandler(ListToolsRequestSchema, async () => { const allTools = [ { name: 'list_projects', description: 'List configured Langfuse projects available to this MCP server.', inputSchema: { type: 'object', properties: {}, }, }, { name: 'project_overview', description: 'Get a summary of total cost, tokens, and traces for a project over a time window.', inputSchema: { type: 'object', properties: { from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, environment: { type: 'string', description: 'Optional environment filter (e.g., "production", "staging")', }, }, required: ['from', 'to'], }, }, { name: 'usage_by_model', description: 'Break down usage and cost by AI model over a time period.', inputSchema: { type: 'object', properties: { from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, environment: { type: 'string', description: 'Optional environment filter', }, limit: { type: 'number', description: 'Maximum number of models to return (default: 20)', }, }, required: ['from', 'to'], }, }, { name: 'usage_by_service', description: 'Analyze usage and cost by service/feature tag over a time period.', inputSchema: { type: 'object', properties: { from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, serviceTagKey: { type: 'string', description: 'Tag key for service identification (default: "service")', }, environment: { type: 'string', description: 'Optional environment filter', }, limit: { type: 'number', description: 'Maximum number of services to return (default: 20)', }, }, required: ['from', 'to'], }, }, { name: 'top_expensive_traces', description: 'Find the most expensive traces by cost over a time period.', inputSchema: { type: 'object', properties: { from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, limit: { type: 'number', description: 'Maximum number of traces to return (default: 10)', }, environment: { type: 'string', description: 'Optional environment filter', }, }, required: ['from', 'to'], }, }, { name: 'get_trace_detail', description: 'Get detailed information about a specific trace including all observations.', inputSchema: { type: 'object', properties: { traceId: { type: 'string', description: 'The trace ID to retrieve', }, }, required: ['traceId'], }, }, // New requested tools { name: 'get_projects', description: 'List available Langfuse projects (alias for list_projects).', inputSchema: { type: 'object', properties: {}, }, }, { name: 'get_metrics', description: 'Query aggregated metrics (costs, tokens, counts) with flexible filtering and dimensions.', inputSchema: { type: 'object', properties: { from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, view: { type: 'string', enum: ['traces', 'observations'], description: 'Data view to query (default: traces)', }, metrics: { type: 'array', items: { type: 'object', properties: { measure: { type: 'string' }, aggregation: { type: 'string' }, }, }, description: 'Metrics to aggregate', }, dimensions: { type: 'array', items: { type: 'object', properties: { field: { type: 'string' }, }, }, description: 'Dimensions to group by', }, environment: { type: 'string', description: 'Optional environment filter', }, }, required: ['from', 'to'], }, }, { name: 'get_traces', description: 'Fetch traces with flexible filtering options.', inputSchema: { type: 'object', properties: { from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, limit: { type: 'number', minimum: 1, maximum: 100, description: 'Maximum number of traces to return (default: 25)', }, orderBy: { type: 'string', enum: ['timestamp', 'totalCost', 'totalTokens'], description: 'Field to order by (default: timestamp)', }, orderDirection: { type: 'string', enum: ['asc', 'desc'], description: 'Order direction (default: desc)', }, userId: { type: 'string', description: 'Filter by user ID', }, name: { type: 'string', description: 'Filter by trace name (substring match)', }, tags: { type: 'array', items: { type: 'string' }, description: 'Filter by tags', }, environment: { type: 'string', description: 'Filter by environment', }, minCost: { type: 'number', description: 'Minimum cost filter', }, maxCost: { type: 'number', description: 'Maximum cost filter', }, }, }, }, { name: 'get_observations', description: 'Get LLM generations/spans with details and filtering.', inputSchema: { type: 'object', properties: { traceId: { type: 'string', description: 'Filter by specific trace ID', }, from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, limit: { type: 'number', minimum: 1, maximum: 100, description: 'Maximum number of observations to return (default: 25)', }, type: { type: 'string', enum: ['GENERATION', 'SPAN', 'EVENT'], description: 'Filter by observation type', }, model: { type: 'string', description: 'Filter by model name (substring match)', }, name: { type: 'string', description: 'Filter by observation name (substring match)', }, level: { type: 'string', enum: ['DEBUG', 'DEFAULT', 'WARNING', 'ERROR'], description: 'Filter by log level', }, }, }, }, { name: 'get_cost_analysis', description: 'Specialized cost breakdowns by model, user, and daily trends.', inputSchema: { type: 'object', properties: { from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, environment: { type: 'string', description: 'Optional environment filter', }, includeModelBreakdown: { type: 'boolean', description: 'Include breakdown by model (default: true)', }, includeUserBreakdown: { type: 'boolean', description: 'Include breakdown by user (default: true)', }, includeDailyBreakdown: { type: 'boolean', description: 'Include daily breakdown (default: true)', }, limit: { type: 'number', minimum: 5, maximum: 100, description: 'Maximum items per breakdown (default: 20)', }, }, required: ['from', 'to'], }, }, { name: 'get_daily_metrics', description: 'Daily usage trends and patterns.', inputSchema: { type: 'object', properties: { from: { type: 'string', format: 'date-time', description: 'Start timestamp (ISO 8601)', }, to: { type: 'string', format: 'date-time', description: 'End timestamp (ISO 8601)', }, environment: { type: 'string', description: 'Optional environment filter', }, fillMissingDays: { type: 'boolean', description: 'Fill missing days with zero values (default: true)', }, }, required: ['from', 'to'], }, }, // Additional API tools { name: 'get_observation_detail', description: 'Get detailed information about a specific observation by ID.', inputSchema: { type: 'object', properties: { observationId: { type: 'string', description: 'The observation ID to retrieve detailed information for', }, }, required: ['observationId'], }, }, { name: 'get_health_status', description: 'Get system health status and availability information.', inputSchema: { type: 'object', properties: {}, }, }, { name: 'list_models', description: 'List all available AI models in the Langfuse project.', inputSchema: { type: 'object', properties: { limit: { type: 'number', description: 'Maximum number of models to return (default: 50)', }, page: { type: 'number', description: 'Page number for pagination', }, }, }, }, { name: 'get_model_detail', description: 'Get detailed information about a specific AI model.', inputSchema: { type: 'object', properties: { modelId: { type: 'string', description: 'The model ID to retrieve detailed information for', }, }, required: ['modelId'], }, }, { name: 'list_prompts', description: 'List all prompt templates in the Langfuse project.', inputSchema: { type: 'object', properties: { limit: { type: 'number', description: 'Maximum number of prompts to return (default: 50)', }, page: { type: 'number', description: 'Page number for pagination', }, name: { type: 'string', description: 'Filter by prompt name (substring match)', }, }, }, }, { name: 'get_prompt_detail', description: 'Get detailed information about a specific prompt template.', inputSchema: { type: 'object', properties: { promptName: { type: 'string', description: 'The prompt name to retrieve', }, version: { type: 'number', description: 'Specific version to retrieve (if not provided, gets latest)', }, label: { type: 'string', description: 'Specific label to retrieve', }, }, required: ['promptName'], }, }, // Dataset management tools { name: 'create_dataset', description: 'Create a new dataset for organizing test data and examples.', inputSchema: { type: 'object', properties: { name: { type: 'string', description: 'Name of the dataset (required)', }, description: { type: 'string', description: 'Optional description of the dataset', }, metadata: { type: 'object', description: 'Optional metadata object for the dataset', }, }, required: ['name'], }, }, { name: 'list_datasets', description: 'List all datasets in the project with pagination support.', inputSchema: { type: 'object', properties: { page: { type: 'number', description: 'Page number for pagination (starts at 1)', }, limit: { type: 'number', description: 'Maximum number of datasets to return (default: 50)', }, }, }, }, { name: 'get_dataset', description: 'Get detailed information about a specific dataset by name.', inputSchema: { type: 'object', properties: { datasetName: { type: 'string', description: 'Name of the dataset to retrieve', }, }, required: ['datasetName'], }, }, { name: 'create_dataset_item', description: 'Add a new item to a dataset with input/output examples.', inputSchema: { type: 'object', properties: { datasetName: { type: 'string', description: 'Name of the dataset to add the item to (required)', }, input: { type: 'object', description: 'Input data for the dataset item', }, expectedOutput: { type: 'object', description: 'Expected output for the dataset item', }, metadata: { type: 'object', description: 'Optional metadata object for the dataset item', }, sourceTraceId: { type: 'string', description: 'Optional trace ID this dataset item is derived from', }, sourceObservationId: { type: 'string', description: 'Optional observation ID this dataset item is derived from', }, status: { type: 'string', enum: ['ACTIVE', 'ARCHIVED'], description: 'Status of the dataset item (default: ACTIVE)', }, }, required: ['datasetName'], }, }, { name: 'list_dataset_items', description: 'List items in datasets with filtering and pagination.', inputSchema: { type: 'object', properties: { datasetName: { type: 'string', description: 'Filter by dataset name', }, sourceTraceId: { type: 'string', description: 'Filter by source trace ID', }, sourceObservationId: { type: 'string', description: 'Filter by source observation ID', }, page: { type: 'number', description: 'Page number for pagination (starts at 1)', }, limit: { type: 'number', description: 'Maximum number of items to return (default: 50)', }, }, }, }, { name: 'get_dataset_item', description: 'Get detailed information about a specific dataset item.', inputSchema: { type: 'object', properties: { itemId: { type: 'string', description: 'ID of the dataset item to retrieve', }, }, required: ['itemId'], }, }, { name: 'delete_dataset_item', description: '⚠️ DESTRUCTIVE: Delete a dataset item permanently from the dataset.', inputSchema: { type: 'object', properties: { itemId: { type: 'string', description: 'ID of the dataset item to delete', }, confirmed: { type: 'boolean', description: 'Set to true to confirm you want to permanently delete this item', }, }, required: ['itemId'], }, }, // Comment management tools { name: 'create_comment', description: 'Create a new comment on traces, observations, sessions, or prompts.', inputSchema: { type: 'object', properties: { objectType: { type: 'string', enum: ['trace', 'observation', 'session', 'prompt'], description: 'The type of object to comment on', }, objectId: { type: 'string', description: 'The unique identifier of the object to comment on', }, content: { type: 'string', description: 'The content/text of the comment', }, authorUserId: { type: 'string', description: 'The user ID of the comment author (optional)', }, }, required: ['objectType', 'objectId', 'content'], }, }, { name: 'list_comments', description: 'List comments with filtering options for objects and users.', inputSchema: { type: 'object', properties: { page: { type: 'number', description: 'Page number for pagination (starts at 1)', }, limit: { type: 'number', minimum: 1, maximum: 100, description: 'Maximum number of comments to return (max 100)', }, objectType: { type: 'string', enum: ['trace', 'observation', 'session', 'prompt'], description: 'Filter comments by object type', }, objectId: { type: 'string', description: 'Filter comments by specific object ID', }, authorUserId: { type: 'string', description: 'Filter comments by author user ID', }, }, }, }, { name: 'get_comment', description: 'Get detailed information about a specific comment.', inputSchema: { type: 'object', properties: { commentId: { type: 'string', description: 'The unique identifier of the comment to retrieve', }, }, required: ['commentId'], }, }, // Write operations with clear prefixing (New in v1.4.0) { name: 'write_create_dataset', description: 'Create a new dataset for organizing test data and examples. This write operation will modify your Langfuse project.', inputSchema: { type: 'object', properties: { name: { type: 'string', description: 'Name of the dataset', }, description: { type: 'string', description: 'Description of the dataset', }, metadata: { type: 'object', description: 'Additional metadata for the dataset', }, }, required: ['name'], }, }, { name: 'write_create_dataset_item', description: 'Add an item to a dataset with input/output examples. This write operation will modify your Langfuse project.', inputSchema: { type: 'object', properties: { datasetName: { type: 'string', description: 'Name of the dataset to add the item to', }, input: { description: 'Input data for the dataset item', }, expectedOutput: { description: 'Expected output for the dataset item', }, sourceTraceId: { type: 'string', description: 'Optional source trace ID to link this item', }, sourceObservationId: { type: 'string', description: 'Optional source observation ID to link this item', }, metadata: { type: 'object', description: 'Additional metadata for the item', }, status: { type: 'string', enum: ['ACTIVE', 'ARCHIVED'], description: 'Status of the dataset item', }, }, required: ['datasetName', 'input'], }, }, { name: 'write_delete_dataset_item', description: '⚠️ DESTRUCTIVE: Permanently delete a dataset item. This write operation cannot be undone.', inputSchema: { type: 'object', properties: { itemId: { type: 'string', description: 'The unique identifier of the dataset item to delete', }, confirmed: { type: 'boolean', description: 'Set to true to confirm you want to permanently delete this item', }, }, required: ['itemId'], }, }, { name: 'write_create_comment', description: 'Create a comment on a trace, observation, session, or prompt. This write operation will modify your Langfuse project.', inputSchema: { type: 'object', properties: { objectType: { type: 'string', enum: ['trace', 'observation', 'session', 'prompt'], description: 'Type of object to comment on', }, objectId: { type: 'string', description: 'ID of the object to comment on', }, content: { type: 'string', description: 'Content of the comment', }, authorUserId: { type: 'string', description: 'Optional author user ID (defaults to API key user)', }, }, required: ['objectType', 'objectId', 'content'], }, }, ]; // Filter tools based on current mode const allowedTools = allTools.filter(tool => isToolAllowed(tool.name, this.modeConfig) ); // Add mode indicators to tool descriptions const enhancedTools = allowedTools.map(tool => { if (isWriteTool(tool.name) && this.mode === 'readwrite') { return { ...tool, description: `🔴 [WRITE] ${tool.description}` }; } return tool; }); console.error(`📊 Exposing ${enhancedTools.length}/${allTools.length} tools in ${this.mode} mode`); return { tools: enhancedTools }; }); // Call tool handler with mode validation this.server.setRequestHandler(CallToolRequestSchema, async (request) => { const startTime = Date.now(); const toolName = request.params.name; try { // Runtime mode validation (defense in depth) if (!isToolAllowed(toolName, this.modeConfig)) { this.auditLogger.logPermissionDenied(toolName, this.mode); throw new Error( `Permission denied: "${toolName}" requires read-write mode. ` + `This server is running in ${this.mode} mode. ` + `Set LANGFUSE_MCP_MODE=readwrite to enable write operations.` ); } // Log start of write operations if (isWriteTool(toolName)) { this.auditLogger.logOperationStart(toolName, request.params.arguments || {}); } switch (toolName) { case 'list_projects': { const args = listProjectsSchema.parse(request.params.arguments); return await listProjects(this.client); } case 'project_overview': { const args = projectOverviewSchema.parse(request.params.arguments); return await projectOverview(this.client, args); } case 'usage_by_model': { const args = usageByModelSchema.parse(request.params.arguments); return await usageByModel(this.client, args); } case 'usage_by_service': { const args = usageByServiceSchema.parse(request.params.arguments); return await usageByService(this.client, args); } case 'top_expensive_traces': { const args = topExpensiveTracesSchema.parse(request.params.arguments); return await topExpensiveTraces(this.client, args); } case 'get_trace_detail': { const args = getTraceDetailSchema.parse(request.params.arguments); return await getTraceDetail(this.client, args); } // New requested tools case 'get_projects': { const args = getProjectsSchema.parse(request.params.arguments); return await getProjects(this.client); } case 'get_metrics': { const args = getMetricsSchema.parse(request.params.arguments); return await getMetrics(this.client, args); } case 'get_traces': { const args = getTracesSchema.parse(request.params.arguments); return await getTraces(this.client, args); } case 'get_observations': { const args = getObservationsSchema.parse(request.params.arguments); return await getObservations(this.client, args); } case 'get_cost_analysis': { const args = getCostAnalysisSchema.parse(request.params.arguments); return await getCostAnalysis(this.client, args); } case 'get_daily_metrics': { const args = getDailyMetricsSchema.parse(request.params.arguments); return await getDailyMetrics(this.client, args); } // Additional API tools case 'get_observation_detail': { const args = getObservationDetailSchema.parse(request.params.arguments); return await getObservationDetail(this.client, args); } case 'get_health_status': { const args = getHealthStatusSchema.parse(request.params.arguments); return await getHealthStatus(this.client, args); } case 'list_models': { const args = listModelsSchema.parse(request.params.arguments); return await listModels(this.client, args); } case 'get_model_detail': { const args = getModelDetailSchema.parse(request.params.arguments); return await getModelDetail(this.client, args); } case 'list_prompts': { const args = listPromptsSchema.parse(request.params.arguments); return await listPrompts(this.client, args); } case 'get_prompt_detail': { const args = getPromptDetailSchema.parse(request.params.arguments); return await getPromptDetail(this.client, args); } // Dataset management tools case 'create_dataset': { const args = createDatasetSchema.parse(request.params.arguments); return await createDataset(this.client, args); } case 'list_datasets': { const args = listDatasetsSchema.parse(request.params.arguments); return await listDatasets(this.client, args); } case 'get_dataset': { const args = getDatasetSchema.parse(request.params.arguments); return await getDataset(this.client, args); } case 'create_dataset_item': { const args = createDatasetItemSchema.parse(request.params.arguments); return await createDatasetItem(this.client, args); } case 'list_dataset_items': { const args = listDatasetItemsSchema.parse(request.params.arguments); return await listDatasetItems(this.client, args); } case 'get_dataset_item': { const args = getDatasetItemSchema.parse(request.params.arguments); return await getDatasetItem(this.client, args); } case 'delete_dataset_item': { const args = deleteDatasetItemSchema.parse(request.params.arguments); // Validate confirmation for destructive operation validateConfirmation(toolName, args.confirmed); return await deleteDatasetItem(this.client, args); } // Comment management tools case 'create_comment': { const args = createCommentSchema.parse(request.params.arguments); return await createComment(this.client, args); } case 'list_comments': { const args = listCommentsSchema.parse(request.params.arguments); return await listComments(this.client, args); } case 'get_comment': { const args = getCommentSchema.parse(request.params.arguments); return await getComment(this.client, args); } // New prefixed write tools (v1.4.0) - with enhanced validation case 'write_create_dataset': { const args = createDatasetSchema.parse(request.params.arguments); return await createDataset(this.client, args); } case 'write_create_dataset_item': { const args = createDatasetItemSchema.parse(request.params.arguments); return await createDatasetItem(this.client, args); } case 'write_delete_dataset_item': { const args = deleteDatasetItemSchema.parse(request.params.arguments); // Validate confirmation for destructive operation validateConfirmation(toolName, args.confirmed); return await deleteDatasetItem(this.client, args); } case 'write_create_comment': { const args = createCommentSchema.parse(request.params.arguments); return await createComment(this.client, args); } default: throw new Error(`Unknown tool: ${toolName}`); } // Log successful write operations const duration = Date.now() - startTime; if (isWriteTool(toolName)) { // Note: We can't access the actual result here without changing the return flow // But we know it succeeded if we reach this point this.auditLogger.logOperationSuccess(toolName, request.params.arguments || {}, undefined, duration); } } catch (error) { const duration = Date.now() - startTime; const errorMessage = error instanceof Error ? error.message : String(error); // Log failed write operations if (isWriteTool(toolName)) { const err = error instanceof Error ? error : new Error(String(error)); this.auditLogger.logOperationError(toolName, request.params.arguments || {}, err, duration); } return { content: [ { type: 'text' as const, text: `Error: ${errorMessage}`, }, ], isError: true, }; } }); } async run(): Promise<void> { const transport = new StdioServerTransport(); await this.server.connect(transport); console.error('Langfuse Analytics MCP server running on stdio'); } } const server = new LangfuseAnalyticsServer(); server.run().catch(console.error);