Web Scraping API Pricing: Costs for 1,000 and 10,000 Pages

👁 113 reads

Web scraping prices are easy to compare incorrectly because providers sell different units. This guide calculates defined 1,000- and 10,000-unit scenarios as of October 2026, using USD prices checked on October 23 and keeping allowances separate from gross usage.

Define the unit before multiplying the price

Start with what the application needs: known pages, discovered sources, structured records, or interactive browser tasks. A search query can return several candidate URLs. A source page can contain several records. A browser session can visit multiple pages. None of those relationships has a universal fixed ratio.

The extraction examples below assume basic operations, no optional enrichments, and the stated plan or PAYG rate. They show gross spending before free allowances unless a row explicitly says otherwise. These are calculations from published tariffs, not measured bills from paid production runs.

For a decision, also estimate the number of accepted results. A response that contains a login notice may consume resources without helping the user. Track retries and discarded results separately so an attractive per-request rate does not hide poor economics for the actual task.

Page extraction: 1,000 and 10,000 scenarios

Firecrawl’s first row deliberately shows the paid-plan bill. For 1,000 basic pages, its free tier may be enough if the account and usage qualify. If you choose Hobby, however, the monthly subscription is $19, even if you use only 1,000 credits. The plan’s full-utilization allocation is $19 divided by 5,000, or $3.80 per 1,000 pages.

For 10,000 basic pages on Hobby, the calculation is $19 plus five additional blocks of 1,000 credits at $5, totaling $44. This assumes no extra output-format charges and does not add a separate free-plan allowance to the paid plan. Annual-billing equivalents are excluded from this monthly comparison.

Tavily’s PAYG credit is $0.008. Basic Extract uses one credit per five successful URL extractions: 1,000 divided by five, multiplied by $0.008, equals $1.60. Basic Crawl adds regular mapping at one credit per ten pages, producing $2.40 for 1,000 pages. Account allowances can reduce actual payment.

Exa Contents charges $1 per 1,000 pages per content type. Requesting more content types changes the calculation. Local Crawl4AI has no per-page library license charge, but that zero does not include computing, engineering, or external services.

Records and platform usage need separate rows

Bright Data Web Scraper API lists $1.50 per 1,000 records on PAYG. That yields gross usage of $1.50 for 1,000 records and $15 for 10,000 records. These are record scenarios, not a claim that each record equals one page. The product also advertises 5,000 free monthly records, subject to its terms.

If that allowance applies to the same workload, 10,000 records would leave 5,000 billable records, or $7.50 at the listed PAYG rate. Confirm eligibility and the specific product in the account before using the net figure. The gross rate is the more portable comparison between workloads.

Apify Starter costs $19 per month and includes $19 of usage. Its free plan lists $5 per month. Actors can charge by events, resource consumption, or both. A correct estimate therefore needs the selected Actor’s tariff and output unit. Dividing the platform subscription by an invented page allowance would create a misleading number.

Search costs are another part of the bill

These figures are gross query costs before allowances. Exa’s starting rate depends on search mode. They do not include every downstream content read or model response. The search API comparison explains how those stages differ.

An assistant that already has a source URL can avoid an unnecessary discovery step. A research task that needs several independent sources may require multiple reads. Build the budget from that workflow rather than assuming that every user question costs one search and one page.

Include the costs that tariffs leave out

Add storage, cleanup, monitoring, and model inference where they apply. For browser agents, include session duration and traffic as well as model calls. Introductory credits may make the first experiment cheap without changing the ongoing operating cost.

• Count billable operations and accepted results separately.
• Record output options that increase credit consumption.
• Keep subscription minimums distinct from full-utilization allocations.
• Recheck rates when the workload or plan changes.

The Firecrawl review details its plan math, and the Crawl4AI tutorial shows what local setup involves. Use the agent data guide to select the required operations before choosing the cheapest advertised unit.

FAQ

Why does 1,000 Firecrawl pages show $19?

That row models purchasing Hobby for the month. The free plan may cover qualifying basic usage, but the paid subscription still has a minimum bill. The $3.80 allocation assumes using all 5,000 included credits.

Which service is cheapest?

There is no universal answer across different units and outputs. Compare a defined workload with the same acceptance criteria, accounting for allowances, options, and operating costs.

Are these paid test results?

No. They are transparent calculations from published prices. No paid production scraping benchmark was run for this article, and the figures do not predict a provider’s success rate.

Does a free local library remove all costs?

No. It removes a per-page library fee. Hardware, maintenance, troubleshooting, and optional external services can still contribute to the total cost of a usable result.

Kinetyra