What is a rate limit?
A rate limit is the most requests an API accepts in a period: per second, per minute or per day. Past it, requests are refused, usually with a 429 “too many requests” response, until the period resets. It is related to concurrency, the number of requests running at the same moment, and vendors often publish one, the other, or both for each plan.
Rate limit across the APIs we track
| API | What the vendor publishes | Read |
|---|---|---|
| AgentQL | Starter is limited to 10 API calls per minute and 5 concurrent remote-browser sessions. | Jul 26, 2026 |
| AgentQL | Professional is limited to 50 API calls per minute and 100 concurrent remote-browser sessions. | Jul 26, 2026 |
| Anchor Browser | Free session creation is limited to 5 per minute, Starter to 25 per minute, and Growth to 100 per minute. | Jul 26, 2026 |
| Brave Search API | The self-serve Search plan publishes a 50-query-per-second rate limit. | Jul 26, 2026 |
| cloro | All endpoints are rate limited to 1,000 requests per second; concurrency limits depend on the plan and exceeding either returns HTTP 429. | Oct 11, 2026 |
| Google Custom Search JSON API | Paid usage is capped at 10,000 queries per day. | Jul 26, 2026 |
| Gumloop | On Pro, an agent interaction that arrives while all 25 concurrency slots are busy is rejected (HTTP 429 through the API); Enterprise queues instead. | Oct 12, 2026 |
| Jina AI Reader and Search Foundation APIs | No-key Reader access is limited to 20 requests per minute; free and paid API keys are shown at 500 requests per minute, while premium keys are shown at 5,000 requests per minute. | Jul 26, 2026 |
| Jina AI Reader and Search Foundation APIs | Search is unavailable without an API key and is shown at 100 requests per minute for free and paid keys. | Jul 26, 2026 |
| Microsoft Bing Search API | The replacement publishes limits of 150 transactions per second and one million transactions per day. | Jul 26, 2026 |
| Parallel Web Systems Search API | Search processors publish a limit of 600 requests per minute. | Jul 26, 2026 |
| Scavio | The free and pay-as-you-go tiers are limited to one request per second. | Sep 2, 2026 |
| Scavio | Request rates are 5 per second on Project, 10 on Bootstrap, 15 on Startup and 50 on Growth. | Oct 9, 2026 |
| ScrapeBadger | Rate limits per plan from the plan table: PAYG 60 requests per minute, Starter and Growth 300, Pro and Scale 1,000. | Oct 12, 2026 |
| ScrapeGraphAI | Starter allows 100 requests per minute, 5 monitors, and 3 concurrent crawls. | Jul 26, 2026 |
| ScrapeGraphAI | Growth allows 500 requests per minute, 25 monitors, and 15 concurrent crawls. | Jul 26, 2026 |
| ScrapeGraphAI | Pro allows 5,000 requests per minute, 100 monitors, and 50 concurrent crawls. | Jul 26, 2026 |
| ScraperAPI | Requests above the concurrency limit receive HTTP 429 and are not queued. | Jul 26, 2026 |
| Scraping Robot | The default customer limit is a soft 100 requests per minute and can be adjusted by contacting the vendor. | Jul 26, 2026 |
| SearchApi.io | The hourly rate limit is up to 20 percent of the plan's monthly searches. | Jul 26, 2026 |
| Serper.dev | Published rate limits range from 50 queries per second on Starter to 300 queries per second on Ultimate. | Jul 26, 2026 |
| serpstack | The paid plan cards state No Rate Limit. | Sep 2, 2026 |
| Steel.dev | Launch allows 10 concurrent sessions, 15-minute maximum sessions, 3 seats, 60 requests per minute, and 7-day retention. | Jul 26, 2026 |
| Steel.dev | Scale allows 100 concurrent sessions, 1-hour maximum sessions, unlimited seats, 600 requests per minute, and 14-day retention. | Jul 26, 2026 |
| Thunderbit | Open API concurrency and rate limits conflict between pages: the product page and Thunderbit's API blog say Starter 30 and Pro 50 concurrent requests; the docs rate-limits page says Free 2 (10 requests per minute), Pro 10 (100 per minute) and Enterprise 50 (1,000 per minute). | Oct 11, 2026 |
| TinyFish | Search is limited to 30 requests per minute and 500 requests per hour. Fetch is limited to 150 URLs per minute and 1,000 URLs per day. These rate limits, not a price, are what bound the free surface. | Sep 1, 2026 |
| ValueSERP | The 25K, 200K, and 1M tiers publish limits of 250, 1,000, and 1,500 requests per minute respectively for real-time requests. | Jul 26, 2026 |
| Zyte API | Standard plans are rate limited to 3,000 requests a minute; Enterprise to 10,000. | Oct 11, 2026 |
Only vendors that publish the term are listed; a vendor missing here has not published it, which is not the same as having none.
In more detail
The vendor examples below were read on the dates given in the text; the table above holds each vendor's current figure.
Why it matters for the bill
A rate limit does not change the price of a request, but it caps how much of a plan you can use in a given time. A plan with a large allowance and a low rate limit may not finish a big crawl inside the month, which pushes you to a bigger plan for its speed rather than its volume. Refused requests are often not charged, but retrying them takes time; the 429 page shows what each vendor’s billing rule says and the concurrency each plan allows.
The table lists the limits each vendor publishes, with the day each was read.