Imported from jeremylongshore/tons-of-skills-marketplace (skills/.curated/figma-rate-limits/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill figma-rate-limits. Copyright stays with the author (MIT).

Figma Rate Limits

Overview

Figma uses a leaky bucket algorithm for rate limiting. When the bucket is full, the API returns 429 with a Retry-After header. Limits vary by plan tier, seat type, and endpoint tier.

Prerequisites

• Figma REST API integration working
• Understanding of async/await patterns

Instructions

Step 1: Understand the Rate Limit Model

Endpoint tiers (limits are per-user, per-minute):

429 response headers:

Step 2: Implement Exponential Backoff

async function figmaFetchWithRetry(
  path: string,
  token: string,
  maxRetries = 5
): Promise<any> {
  for (let attempt = 0; attempt <= maxRetries; attempt++) {
    const res = await fetch(`https://api.figma.com${path}`, {
      headers: { 'X-Figma-Token': token },
    });

    if (res.status === 429) {
      const retryAfter = parseInt(res.headers.get('Retry-After') || '60');
      const limitType = res.headers.get('X-Figma-Rate-Limit-Type') || 'unknown';

      if (attempt === maxRetries) {
        throw new Error(`Rate limited after ${maxRetries} retries (${limitType})`);
      }

      // Use the Retry-After header -- Figma tells you exactly how long to wait
      const jitter = Math.random() * 1000;
      const delay = retryAfter * 1000 + jitter;
      console.warn(`429 (${limitType}). Waiting ${(delay/1000).toFixed(1)}s (attempt ${attempt + 1})`);
      await new Promise(r => setTimeout(r, delay));
      continue;
    }

    if (res.status >= 500 && attempt < maxRetries) {
      // Server errors: exponential backoff without Retry-After
      const delay = Math.min(1000 * Math.pow(2, attempt), 30000);
      await new Promise(r => setTimeout(r, delay));
      continue;
    }

    if (!res.ok) {
      throw new Error(`Figma API error: ${res.status} ${await res.text()}`);
    }

    return res.json();
  }
}

Step 3: Request Queue with Concurrency Control

import PQueue from 'p-queue';

// Limit concurrent requests to avoid bursting the bucket
const figmaQueue = new PQueue({
  concurrency: 3,       // max 3 parallel requests
  interval: 1000,       // per second
  intervalCap: 5,       // max 5 requests per second
});

async function queuedFigmaRequest<T>(
  path: string,
  token: string
): Promise<T> {
  return figmaQueue.add(() => figmaFetchWithRetry(path, token));
}

// Usage -- all requests are automatically queued and throttled
const [file, comments, images] = await Promise.all([
  queuedFigmaRequest(`/v1/files/${fileKey}`, token),
  queuedFigmaRequest(`/v1/files/${fileKey}/comments`, token),
  queuedFigmaRequest(`/v1/images/${fileKey}?ids=0:1&format=svg`, token),
]);

Step 4: Rate Limit Monitor

class FigmaRateLimitMonitor {
  private requestLog: number[] = [];
  private windowMs = 60_000; // 1 minute window

  recordRequest() {
    this.requestLog.push(Date.now());
    // Trim old entries
    const cutoff = Date.now() - this.windowMs;
    this.requestLog = this.requestLog.filter(t => t > cutoff);
  }

  getRequestsInWindow(): number {
    const cutoff = Date.now() - this.windowMs;
    return this.requestLog.filter(t => t > cutoff).length;
  }

  shouldThrottle(safetyMargin = 0.8): boolean {
    // If we've used 80% of a conservative estimate, slow down
    const estimatedLimit = 30; // Conservative estimate
    return this.getRequestsInWindow() > estimatedLimit * safetyMargin;
  }
}

const monitor = new FigmaRateLimitMonitor();

// Wrap every request
async function monitoredFigmaFetch(path: string, token: string) {
  if (monitor.shouldThrottle()) {
    console.warn('Approaching rate limit, adding delay');
    await new Promise(r => setTimeout(r, 2000));
  }
  monitor.recordRequest();
  return figmaFetchWithRetry(path, token);
}

Step 5: Batch Node Requests

// Instead of N individual /v1/files/:key/nodes requests,
// batch node IDs into fewer requests
async function batchFetchNodes(
  fileKey: string,
  nodeIds: string[],
  batchSize = 50,
  token: string
) {
  const results: Record<string, any> = {};

  for (let i = 0; i < nodeIds.length; i += batchSize) {
    const batch = nodeIds.slice(i, i + batchSize);
    const ids = encodeURIComponent(batch.join(','));
    const data = await queuedFigmaRequest(
      `/v1/files/${fileKey}/nodes?ids=${ids}`,
      token
    );
    Object.assign(results, data.nodes);
  }

  return results;
}

Output

• Automatic retry with Retry-After header compliance
• Request queue preventing burst overload
• Rate limit monitoring with proactive throttling
• Batch operations reducing total request count

Error Handling

Examples

Reproduce a 429 and read the headers that drive every pattern in this skill (Step 1):

for i in $(seq 1 60); do
  curl -s -o /dev/null -D - -H "X-Figma-Token: ${FIGMA_PAT}" \
    "https://api.figma.com/v1/files/${FIGMA_FILE_KEY}?depth=1" | /usr/bin/grep -iE '^(HTTP|retry-after)'
done | sort | uniq -c

  54 HTTP/2 200
   6 HTTP/2 429
   6 retry-after: 30

Collapse N per-node calls into one batched request (Step 5) — the single biggest budget win:

curl -s -H "X-Figma-Token: ${FIGMA_PAT}" \
  "https://api.figma.com/v1/files/${FIGMA_FILE_KEY}/nodes?ids=1:2,1:5,1:9,2:14" | jq '.nodes | keys'

Backoff implementation and the queue: references/implement-exponential-backoff.md, references/request-queue-with-concurrency-control.md.

Resources

• Figma Rate Limits Documentation
• What if I'm rate-limited?
• p-queue Documentation

Next Steps

For security configuration, see figma-security-basics.