Back to Documentation

Cost and Performance Metadata

Rich metadata that helps agents make intelligent decisions about cost, latency, and resource usage.

Complete Tool Transparency

Every MCPify tool exposes comprehensive metadata so agents can optimize their strategies:

{
  "tool": "search_products",
  "metadata": {
    "cost": {
      "per_call_usd": 0.0012,
      "token_estimate": {
        "input": 150,
        "output": 800
      }
    },
    "performance": {
      "latency_ms": {
        "p50": 127,
        "p95": 312,
        "p99": 589
      },
      "throughput_rps": 100
    },
    "limits": {
      "rate_limit_rpm": 600,
      "max_page_size": 100,
      "max_results": 10000
    },
    "cache": {
      "ttl_seconds": 300,
      "hit_rate": 0.87
    }
  }
}

Cost Optimization Signals

Per-Call Costs

  • • API request cost
  • • Data transfer fees
  • • Processing charges
  • • Token consumption

Cost Comparison

  • • Alternative tool costs
  • • Cached vs live costs
  • • Batch vs individual
  • • Peak vs off-peak pricing

Latency Profiles

Detailed latency distributions help agents plan for different scenarios:

P50 (Median)

127ms

Most requests

P95

312ms

Slower requests

P99

589ms

Worst case

Factors affecting latency: Payload size, query complexity, cache status, current load, geographic distance

Resource Utilization

Token Consumption

Input tokens:~150
Output tokens:~800

Memory Usage

Typical response size: 2-5KB • Max response: 100KB

Concurrency

Max parallel calls: 10 • Recommended: 5

Rate Limit Awareness

Dynamic Rate Limit Info

{
  "rate_limits": {
    "tier": "standard",
    "limits": {
      "requests_per_minute": 600,
      "requests_per_day": 100000,
      "concurrent_requests": 10
    },
    "current": {
      "rpm_remaining": 547,
      "rpd_remaining": 87234,
      "reset_at": "2025-08-25T10:30:00Z"
    },
    "headers": {
      "X-RateLimit-Limit": "600",
      "X-RateLimit-Remaining": "547",
      "X-RateLimit-Reset": "1724584200"
    }
  }
}

Using Metadata for Decisions

Agent Decision Example

// Agent evaluates two options:

Option A: Live search
- Cost: $0.0012 per call
- Latency: 127ms (p50)
- Freshness: Real-time

Option B: Cached search
- Cost: $0.00001 (token processing only)
- Latency: 3ms
- Freshness: 5 minutes old

Decision: Use cached data (freshness acceptable for use case)

Metadata Configuration

{
  "metadata_config": {
    "expose_costs": true,
    "include_latency_percentiles": ["p50", "p95", "p99"],
    "track_token_usage": true,
    "rate_limit_headers": true,
    "cache_transparency": "full",
    "cost_units": "usd",
    "latency_units": "milliseconds",
    "update_frequency": "real-time"
  }
}

Make every API call count

With full cost and performance transparency, your agents can optimize for any objective.

Get StartedView Pricing