Skip to main content
Glama

watch

Monitor project files for changes and automatically revalidate contracts to detect schema mismatches between data producers and consumers, preventing runtime errors during development.

Instructions

Watch project files for changes and auto-revalidate contracts. Actions: start (begin watching), stop (end watching), status (check state), poll (get pending events).

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
projectDirYesRoot directory with .trace-mcp config
actionNoWatch action (default: start)

Implementation Reference

  • Handler for the 'watch' MCP tool. Parses input, loads TraceProject, retrieves/manages TraceWatcher singleton, and dispatches to start/stop/status/poll actions using watcher methods.
    case 'watch': {
      const input = WatchInput.parse(args);
      const action = input.action || 'start';
      log(`Watch action: ${action} for ${input.projectDir}`);
      
      const project = loadProject(input.projectDir);
      
      if (!project.exists()) {
        throw new Error(`No trace project found at ${input.projectDir}. Run init_project first.`);
      }
      
      const watcher = getWatcher(project);
      
      switch (action) {
        case 'start': {
          // Collect events during startup
          const events: WatchEvent[] = [];
          const eventHandler = (event: WatchEvent) => events.push(event);
          watcher.on('watch-event', eventHandler);
          
          await watcher.start();
          
          // Wait a moment for initial validation
          await new Promise(resolve => setTimeout(resolve, 100));
          watcher.off('watch-event', eventHandler);
          
          return {
            content: [{
              type: 'text',
              text: JSON.stringify({
                success: true,
                action: 'started',
                projectDir: input.projectDir,
                status: watcher.getStatus(),
                events,
              }, null, 2),
            }],
          };
        }
        
        case 'stop': {
          await stopWatcher(project);
          return {
            content: [{
              type: 'text',
              text: JSON.stringify({
                success: true,
                action: 'stopped',
                projectDir: input.projectDir,
              }, null, 2),
            }],
          };
        }
        
        case 'status': {
          return {
            content: [{
              type: 'text',
              text: JSON.stringify({
                success: true,
                action: 'status',
                projectDir: input.projectDir,
                status: watcher.getStatus(),
                activeWatchers: listActiveWatchers(),
              }, null, 2),
            }],
          };
        }
        
        case 'poll': {
          // For polling, collect recent events
          const status = watcher.getStatus();
          return {
            content: [{
              type: 'text',
              text: JSON.stringify({
                success: true,
                action: 'poll',
                projectDir: input.projectDir,
                status,
                lastResult: status.lastResult,
              }, null, 2),
            }],
          };
        }
        
        default:
          throw new Error(`Unknown watch action: ${action}`);
      }
    }
  • Zod input schema for the 'watch' tool defining projectDir and optional action (start/stop/status/poll). Used for validation in handler.
    const WatchInput = z.object({
      projectDir: z.string().describe('Root directory with .trace-mcp config'),
      action: z.enum(['start', 'stop', 'status', 'poll']).default('start').describe('Watch action to perform'),
    });
  • src/index.ts:253-263 (registration)
    MCP tool registration for 'watch' in server listTools response, including name, description, and inputSchema matching WatchInput.
      name: 'watch',
      description: 'Watch project files for changes and auto-revalidate contracts. Actions: start (begin watching), stop (end watching), status (check state), poll (get pending events).',
      inputSchema: {
        type: 'object',
        properties: {
          projectDir: { type: 'string', description: 'Root directory with .trace-mcp config' },
          action: { type: 'string', enum: ['start', 'stop', 'status', 'poll'], description: 'Watch action (default: start)' },
        },
        required: ['projectDir'],
      },
    },
  • Core TraceWatcher class that implements file watching with chokidar, debounced revalidation (extract/trace/compare), event emission. Used by 'watch' tool handler via getWatcher() singleton.
    export class TraceWatcher extends EventEmitter {
      private project: TraceProject;
      private cache: SchemaCache;
      private producerWatcher: FSWatcher | null = null;
      private consumerWatcher: FSWatcher | null = null;
      private debounceTimer: NodeJS.Timeout | null = null;
      private pendingChanges: Map<string, FileChangeData> = new Map();
      private isRunning = false;
      private lastResult: TraceResult | null = null;
    
      constructor(project: TraceProject) {
        super();
        this.project = project;
        this.cache = new SchemaCache(project);
      }
    
      // --------------------------------------------------------------------------
      // Lifecycle
      // --------------------------------------------------------------------------
    
      /**
       * Start watching files
       */
      async start(): Promise<void> {
        if (this.isRunning) {
          throw new Error('Watcher is already running');
        }
    
        const config = this.project.config;
        this.isRunning = true;
    
        // Build glob patterns for producer
        const producerPatterns = config.producer.include.map(
          p => join(this.project.producerPath, p)
        );
        const producerIgnore = config.producer.exclude.map(
          p => join(this.project.producerPath, p)
        );
    
        // Build glob patterns for consumer  
        const consumerPatterns = config.consumer.include.map(
          p => join(this.project.consumerPath, p)
        );
        const consumerIgnore = config.consumer.exclude.map(
          p => join(this.project.consumerPath, p)
        );
    
        // Start producer watcher
        this.producerWatcher = chokidar.watch(producerPatterns, {
          ignored: producerIgnore,
          persistent: true,
          ignoreInitial: true,
        });
    
        this.producerWatcher.on('add', (path: string) => this.onFileChange(path, 'producer', 'add'));
        this.producerWatcher.on('change', (path: string) => this.onFileChange(path, 'producer', 'change'));
        this.producerWatcher.on('unlink', (path: string) => this.onFileChange(path, 'producer', 'unlink'));
        this.producerWatcher.on('error', (err: unknown) => this.onError(err instanceof Error ? err : new Error(String(err))));
    
        // Start consumer watcher
        this.consumerWatcher = chokidar.watch(consumerPatterns, {
          ignored: consumerIgnore,
          persistent: true,
          ignoreInitial: true,
        });
    
        this.consumerWatcher.on('add', (path: string) => this.onFileChange(path, 'consumer', 'add'));
        this.consumerWatcher.on('change', (path: string) => this.onFileChange(path, 'consumer', 'change'));
        this.consumerWatcher.on('unlink', (path: string) => this.onFileChange(path, 'consumer', 'unlink'));
        this.consumerWatcher.on('error', (err: unknown) => this.onError(err instanceof Error ? err : new Error(String(err))));
    
        // Initial validation
        await this.runValidation('initial');
    
        this.emitEvent('ready', {
          producerPath: this.project.producerPath,
          consumerPath: this.project.consumerPath,
        });
      }
    
      /**
       * Stop watching files
       */
      async stop(): Promise<void> {
        if (!this.isRunning) return;
    
        if (this.debounceTimer) {
          clearTimeout(this.debounceTimer);
          this.debounceTimer = null;
        }
    
        if (this.producerWatcher) {
          await this.producerWatcher.close();
          this.producerWatcher = null;
        }
    
        if (this.consumerWatcher) {
          await this.consumerWatcher.close();
          this.consumerWatcher = null;
        }
    
        this.isRunning = false;
        this.emitEvent('stopped', {});
      }
    
      // --------------------------------------------------------------------------
      // Event Handlers
      // --------------------------------------------------------------------------
    
      private onFileChange(
        filePath: string,
        side: 'producer' | 'consumer',
        changeType: 'add' | 'change' | 'unlink'
      ): void {
        const relPath = relative(this.project.rootDir, filePath);
        
        this.emitEvent('file_changed', {
          file: relPath,
          side,
          changeType,
        } as FileChangeData);
    
        // Accumulate changes
        this.pendingChanges.set(filePath, { file: relPath, side, changeType });
    
        // Invalidate cache for this file
        this.cache.invalidateFiles([filePath]);
    
        // Debounce revalidation
        if (this.debounceTimer) {
          clearTimeout(this.debounceTimer);
        }
    
        const debounceMs = this.project.config.options.debounceMs ?? 300;
        this.debounceTimer = setTimeout(() => {
          this.processPendingChanges();
        }, debounceMs);
      }
    
      private onError(error: Error): void {
        this.emitEvent('error', {
          message: error.message,
          stack: error.stack,
        });
      }
    
      // --------------------------------------------------------------------------
      // Validation
      // --------------------------------------------------------------------------
    
      private async processPendingChanges(): Promise<void> {
        const changes = Array.from(this.pendingChanges.values());
        this.pendingChanges.clear();
    
        if (changes.length === 0) return;
    
        const triggeredBy = changes.map(c => c.file).join(', ');
        await this.runValidation(triggeredBy);
      }
    
      private async runValidation(triggeredBy: string): Promise<void> {
        this.emitEvent('revalidating', { triggeredBy });
    
        try {
          // Extract and trace using configured languages
          const producers = await extractProducerSchemas({
            rootDir: this.project.producerPath,
            language: this.project.config.producer.language,
            include: this.project.config.producer.include,
            exclude: this.project.config.producer.exclude,
          });
    
          const consumers = await traceConsumerUsage({
            rootDir: this.project.consumerPath,
            language: this.project.config.consumer.language,
            include: this.project.config.consumer.include,
            exclude: this.project.config.consumer.exclude,
          });
    
          // Compare
          const result = compareSchemas(producers, consumers, {
            strict: this.project.config.options.strict,
            direction: this.project.config.options.direction,
          });
    
          this.lastResult = result;
    
          this.emitEvent('result', {
            result,
            triggeredBy,
            cached: false,
          } as ResultData);
    
        } catch (error) {
          const message = error instanceof Error ? error.message : String(error);
          this.emitEvent('error', { message, triggeredBy });
        }
      }
    
      // --------------------------------------------------------------------------
      // Utility
      // --------------------------------------------------------------------------
    
      private emitEvent(type: WatchEventType, data: unknown): void {
        const event: WatchEvent = {
          type,
          timestamp: new Date().toISOString(),
          data,
        };
        this.emit('watch-event', event);
      }
    
      /**
       * Get the last validation result
       */
      getLastResult(): TraceResult | null {
        return this.lastResult;
      }
    
      /**
       * Get current watcher status
       */
      getStatus(): {
        running: boolean;
        lastResult: TraceResult | null;
        pendingChanges: number;
      } {
        return {
          running: this.isRunning,
          lastResult: this.lastResult,
          pendingChanges: this.pendingChanges.size,
        };
      }
    
      /**
       * Force a revalidation
       */
      async forceRevalidation(): Promise<TraceResult> {
        await this.runValidation('manual');
        return this.lastResult!;
      }
    }
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations are provided, so the description carries the full burden. It mentions actions and auto-revalidation, but lacks details on behavioral traits: it doesn't specify what 'auto-revalidate contracts' entails (e.g., triggers, effects), whether the tool runs persistently, permission requirements, error handling, or rate limits. For a tool with potential side effects (auto-revalidation), this is a significant gap in disclosure.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is appropriately sized and front-loaded: the first sentence states the core purpose, and the second lists actions efficiently. Every sentence adds value, though it could be slightly more structured (e.g., separating purpose from action details). No wasted words, but not perfectly optimized.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness2/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity (managing file watching with auto-revalidation), no annotations, no output schema, and 2 parameters, the description is incomplete. It doesn't explain what the tool returns (e.g., status output, event details), how auto-revalidation works, or potential side effects. For a tool that likely involves ongoing processes and changes, more context is needed to be fully helpful.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the schema already documents both parameters (projectDir and action with enum). The description adds minimal value beyond the schema by listing the action values, but doesn't explain parameter semantics like what 'projectDir' contains or how actions interact (e.g., if 'start' initiates watching). Baseline 3 is appropriate as the schema does most of the work.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's purpose: 'Watch project files for changes and auto-revalidate contracts.' It specifies the verb ('watch'), resource ('project files'), and outcome ('auto-revalidate contracts'), distinguishing it from siblings like trace_file or get_project_status. However, it doesn't explicitly differentiate from all siblings (e.g., compare might also involve file monitoring).

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description implies usage through the listed actions (start, stop, status, poll), suggesting it's for managing a file-watching process. It doesn't provide explicit when-to-use guidance versus alternatives (e.g., when to use watch vs. trace_file for file-related tasks) or prerequisites, leaving usage context somewhat inferred rather than stated.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server

Other Tools

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/Mnehmos/trace-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server