Performance Matters

Slow agents frustrate users. This article covers techniques to make your OpenClaw agent fast and efficient.

Response Time Optimization

Prompt Engineering

# Keep system prompts concise
# BAD: 5000+ tokens of context
# GOOD: Essential instructions only

## Core Instructions (keep under 500 tokens)
- Be concise
- Confirm before destructive actions
- Use tools when appropriate

Context Window Management

{
  memory: {
    maxContextTokens: 4000,
    summarizeAfter: 10, // messages
    pruneStrategy: 'oldest-first'
  }
}

Tool Execution Speed

Parallel Execution

// Execute independent tools in parallel
const results = await Promise.all([
  tools.fetchEmail(),
  tools.checkCalendar(),
  tools.getWeather()
]);

// Instead of sequential
// const email = await tools.fetchEmail();
// const calendar = await tools.checkCalendar();

Caching

const cache = new LRUCache({ max: 1000 });

async function fetchWithCache(key, fetcher) {
  if (cache.has(key)) {
    return cache.get(key);
  }
  const result = await fetcher();
  cache.set(key, result, { ttl: 300000 }); // 5 min
  return result;
}

Resource Management

Memory Optimization

{
  resources: {
    maxMessageHistory: 50,
    compressOldMessages: true,
    cleanupInterval: 3600000 // 1 hour
  }
}

Connection Pooling

// Reuse HTTP connections
const agent = new Agent({
  keepAlive: true,
  maxSockets: 10
});

Monitoring Performance

// Track response times
const start = Date.now();
const response = await agent.run(input);
const duration = Date.now() - start;

metrics.record('response_time', duration);
if (duration > 5000) {
  logger.warn('Slow response', { duration, input });
}

Conclusion

Performance optimization ensures your agent remains responsive and efficient under real-world loads.