Alchemy Cost Tuning

Alchemy Cost Tuning

Overview

Alchemy pricing is based on Compute Units (CU). Different API methods have different CU costs. Optimize by caching, batching, choosing cheaper methods, and right-sizing your plan.

Plan Comparison

CU Cost Reference (Top Methods)

Prerequisites

• An approved observation window with aggregate method counts and latency; do not record end-user wallet data merely to estimate API usage.
• Current plan and compute-unit limits confirmed in the organization’s Alchemy account, since commercial terms and method costs may change.
• A cache-invalidation policy that distinguishes safe metadata caching from time-sensitive balance, block, and transaction data.

Instructions

Step 1: CU Usage Monitor

// src/cost/cu-monitor.ts
const CU_COSTS: Record<string, number> = {
  'eth_blockNumber': 10, 'eth_getBalance': 19, 'eth_call': 26,
  'getTokenBalances': 50, 'getNftsForOwner': 50, 'getTokenMetadata': 50,
  'getAssetTransfers': 150, 'getNftMetadataBatch': 50,
};

class CuMonitor {
  private usage: Array<{ method: string; cu: number; timestamp: number }> = [];

  record(method: string): void {
    this.usage.push({ method, cu: CU_COSTS[method] || 26, timestamp: Date.now() });
  }

  getHourlyReport(): { totalCu: number; byMethod: Record<string, number> } {
    const cutoff = Date.now() - 3600000;
    const recent = this.usage.filter(u => u.timestamp > cutoff);
    const byMethod: Record<string, number> = {};
    let totalCu = 0;

    for (const u of recent) {
      byMethod[u.method] = (byMethod[u.method] || 0) + u.cu;
      totalCu += u.cu;
    }

    return { totalCu, byMethod };
  }

  getMonthlyProjection(): { projectedMonthly: number; planRecommendation: string } {
    const hourly = this.getHourlyReport();
    const projectedMonthly = hourly.totalCu * 24 * 30;

    let recommendation = 'Free';
    if (projectedMonthly > 300_000_000) recommendation = 'Growth';
    if (projectedMonthly > 1_200_000_000) recommendation = 'Scale';

    return { projectedMonthly, planRecommendation: recommendation };
  }
}

export { CuMonitor };

Step 2: Cost-Optimized Client Wrapper

// src/cost/optimized-client.ts
import { Alchemy, Network } from 'alchemy-sdk';

const cache = new Map<string, { data: any; expiry: number }>();

// Cache token metadata aggressively (rarely changes)
async function getTokenMetadataCached(alchemy: Alchemy, contract: string) {
  const key = `metadata:${contract}`;
  const cached = cache.get(key);
  if (cached && cached.expiry > Date.now()) return cached.data;

  const data = await alchemy.core.getTokenMetadata(contract);
  cache.set(key, { data, expiry: Date.now() + 86400000 }); // 24h cache
  return data;
}

// Use batch instead of individual NFT metadata calls
// 1 batch call (50 CU) vs 100 individual calls (5000 CU)
async function getNftMetadataOptimized(
  alchemy: Alchemy,
  tokens: Array<{ contractAddress: string; tokenId: string }>
) {
  const BATCH_SIZE = 100;
  const results = [];
  for (let i = 0; i < tokens.length; i += BATCH_SIZE) {
    const batch = tokens.slice(i, i + BATCH_SIZE);
    const batchResults = await alchemy.nft.getNftMetadataBatch(batch);
    results.push(...batchResults);
  }
  return results;
}

Step 3: Free Tier Optimization Checklist

// For staying within Free tier (330 CU/sec, 300M CU/month):
// 1. Cache eth_blockNumber (saves 10 CU per redundant call)
// 2. Cache token metadata (saves 50 CU per redundant call)
// 3. Use getNftMetadataBatch instead of getNftMetadata (100x savings)
// 4. Avoid getAssetTransfers loops (150 CU each — cache results)
// 5. Use WebSockets instead of polling (one connection vs repeated calls)
// 6. Rate-limit user-facing endpoints to prevent CU bursts

Output

• CU usage monitor with hourly reports and plan projections
• Cost-optimized client with aggressive caching
• Batch operations reducing CU consumption by 100x
• Free tier optimization checklist

Examples

Collect one hour of aggregate development traffic, feed the counts into CuMonitor, and identify the three methods responsible for the highest CU projection. Add a 24-hour metadata cache and batched NFT metadata requests in the test environment, then compare request counts and response correctness against uncached fixtures. Promote only after cache hits never serve stale time-sensitive balance or transfer data. If observed use approaches the account limit or the monitor’s input is incomplete, throttle the noncritical feature and obtain an updated account-limit report rather than guessing a budget or silently dropping user-facing requests.

Error Handling

Resources

• Alchemy Pricing
• Alchemy Compute Units

Next Steps

For architecture design, see alchemy-reference-architecture.