Free tiers that are actually free

A free tier is useful only when its recurring allowance can perform the job. This page therefore ranks renewable allowances for the stated workload, not the largest-looking number in a pricing card.

What counts as a free tier

The recurring table includes an allowance only when the source says it renews on a stated cadence or remains available under a continuing free plan. A one-time signup balance belongs on the free-trials page. When a vendor does not publish whether an allowance renews, rolls over, expires, or requires a card, the field remains “not published.” Silence is not converted into a favorable rule.

Units stay separate until the dataset has enough published information to convert them. A request, delivered result, record, search, credit, browser minute, and account credit do not buy the same work. Bright Data illustrates the problem even within one vendor: Web Unlocker has a request allowance, while Web Scraper API has a separate record allowance, and the two products use different outputs (Web Unlocker pricing and Web Scraper API pricing, retrieved 2026-07-26). Adding those headline balances together would produce a larger number and a worse answer.

Browserless uses Units across browser time, residential traffic, and CAPTCHA attempts. Started time blocks, traffic, and attempts can draw from the same balance in different ways (pricing and unit documentation, retrieved 2026-07-26). A free balance that covers many simple requests may cover far fewer sessions with added traffic or interaction. The ranking applies the published billing rule for the selected job instead of treating every unit as one page.

Which free tier wins which job

For a short basic-fetch workload, a recurring request or result allowance can lead when the required volume fits inside it. For a rendered workload, a credit tier can lead only after the rendering multiplier is applied. For an interactive job, a browser-time allowance belongs in the eligible set while a response-only API may not. For search, a search-specific allowance can be relevant, but it must not win a general page-scraping comparison merely because its calls look inexpensive.

Firecrawl publishes several free meters with different cadences, including plan credits, browser time, and Agent runs (pricing, retrieved 2026-07-25). The correct winner depends on which of those products the job actually uses. ScraperAPI likewise distinguishes an ongoing free plan from its time-limited trial, and publishes separate request costs for rendering, premium routes, and named targets (pricing and credit rules, retrieved 2026-07-26). Jina Reader presents another shape: access without a key is controlled by a request-rate limit rather than a published monthly no-key allowance (Reader access, retrieved 2026-07-25). A rate limit answers “how quickly,” not “how much renewable monthly capacity,” so it should not be silently converted into a monthly tier.

The live first row is therefore the free tier that covers the selected work under published rules. It is not a permanent declaration that one vendor has the best free plan.

Exclusions and uncertainty

Products that cannot perform the selected job are not ranked, even if they offer a generous allowance for a different product. A browser allowance does not become a page API, and a search allowance does not become a general scraper.

If the public record supports only a minimum bill or leaves a required meter unresolved, the result is a lower bound and is excluded from cheapest claims. If the engine must assume that an unpublished rendering, routing, or output modifier is included, the result is labeled assumption-backed. That result can help form a vendor question, but it is an answer rather than evidence.

The ranking also cannot say that the free tier is sufficient for production. It does not measure success rate, speed, reliability, output quality, target permission, retention needs, support, or the operational cost of a low concurrency limit. This site runs no benchmarks. Check the card requirement, renewal rule, rollover rule, feature locks, concurrency, and billable failure definition before relying on any free allowance, then rerun the workload against the paid offer needed after the allowance ends.