菜单
Handle Anthropic rate limits — understand tiers, implement backoff, Use when working with rate-limits patterns. optimize throughput, and monitor usage. Trigger with "anthropic rate limit", "claude 429", "anthropic throttling", "anthropic usage limits", "claude tokens per minute".
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/ComeOnOliver/skillshub --skill clade-rate-limits命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | clade-rate-limits |
| description | Handle Anthropic rate limits — understand tiers, implement backoff, Use when working with rate-limits patterns. optimize throughput, and monitor usage. Trigger with "anthropic rate limit", "claude 429", "anthropic throttling", "anthropic usage limits", "claude tokens per minute". |
| allowed-tools | Read, Write, Edit |
| version | 1.0.0 |
| license | MIT |
| author | Jeremy Longshore <jeremy@intentsolutions.io> |
| compatible-with | claude-code |
| tags | ["saas","anthropic","claude","rate-limits"] |
Anthropic enforces three types of limits: requests per minute (RPM), input tokens per minute (TPM), and output tokens per minute. Limits depend on your spend tier.
| Tier | Qualification | RPM | Input TPM | Output TPM |
|---|---|---|---|---|
| Tier 1 | Free | 50 | 40,000 | 8,000 |
| Tier 2 | $40+ spend | 1,000 | 80,000 | 16,000 |
| Tier 3 | $200+ spend | 2,000 | 160,000 | 32,000 |
| Tier 4 | $400+ spend | 4,000 | 400,000 | 80,000 |
| Scale | Custom | Custom | Custom | Custom |
Check your tier: console.anthropic.com → Settings → Limits
Every API response includes rate limit headers:
claude-ratelimit-requests-limit: 1000
claude-ratelimit-requests-remaining: 998
claude-ratelimit-requests-reset: 2025-01-01T00:01:00Z
claude-ratelimit-tokens-limit: 80000
claude-ratelimit-tokens-remaining: 79500
claude-ratelimit-tokens-reset: 2025-01-01T00:01:00Z
retry-after: 5
The SDK automatically retries 429 and 529 errors with exponential backoff:
import Anthropic from '@claude-ai/sdk';
const client = new Anthropic({
maxRetries: 3, // default: 2. Set to 0 to disable.
});
async function callWithBackoff(params: Anthropic.MessageCreateParams, maxRetries = 5) {
for (let attempt = 0; attempt < maxRetries; attempt++) {
try {
return await client.messages.create(params);
} catch (err) {
if (err instanceof Anthropic.RateLimitError) {
const retryAfter = Number(err.headers?.['retry-after'] || 2 ** attempt);
const jitter = Math.random() * 1000;
console.log(`Rate limited. Retry in ${retryAfter}s (attempt ${attempt + 1})`);
await new Promise(r => setTimeout(r, retryAfter * 1000 + jitter));
} else {
throw err;
}
}
}
throw new Error('Exceeded max retries');
}
| Strategy | Impact |
|---|---|
| Use Message Batches API | Bypasses rate limits entirely (async, 24h SLA) |
| Use prompt caching | Cached tokens don't count toward input TPM |
| Use smaller models for simple tasks | Lower token counts = more requests per minute |
Pre-count tokens with countTokens | Avoid wasted requests that will fail |
| Queue and batch requests | Smooth out bursts |
// Count before sending — avoid burning RPM on requests that'll fail
const count = await client.messages.countTokens({
model: 'claude-sonnet-4-20250514',
messages,
system: systemPrompt,
});
console.log(`This request will use ${count.input_tokens} input tokens`);
import anthropic
import time
client = anthropic.Anthropic(max_retries=5)
# Or manual handling:
try:
message = client.messages.create(...)
except anthropic.RateLimitError as e:
retry_after = float(e.response.headers.get("retry-after", 5))
time.sleep(retry_after)
maxRetries setting| Error | Cause | Solution |
|---|---|---|
| API Error | Check error type and status code | See clade-common-errors |
See Rate Limit Tiers table, Response Headers section, Built-In SDK Retries, Custom Backoff implementation, and Throughput Optimization strategies above.
See clade-cost-tuning for cost optimization strategies.
clade-install-authEach section contains production-ready code examples. Copy and adapt them to your use case.
Integrate the patterns that match your requirements. Test each change individually.
Run your test suite to confirm the integration works correctly.