Scraping API PricesAll products / AgentQL
AgentQL pricing
How AgentQL bills API extraction and remote browser use, what its public pricing leaves open, and which teams should shortlist it.
published session-time meter only
Fixed for this page; volume is the control.Every available option was priced in advance from published vendor rules. Changing a control only switches between those checked options; nothing is sent anywhere and no account is needed.
Your job
What AgentQL counts
What you pay
Successes you want
200 browser hours
Your stated workload. Every vendor on this site is priced against this same number.
- ÷ 1
Attempts you are billed for
200 browser hours
AgentQL bills failed fetches. This workload assumes none fail, so nothing is added here.
- × 1
What the meter counts
200 browser_hours
One attempt is one billed unit for this workload.
- =
Cash per month
$3.00 /mo
Starter. Cheapest of every eligible purchase option, not the first plan that fits.
Bars share one linear scale across the 3 counted stages, so a multiplier shows as length. Units differ by stage and are labelled. Cash is not on that scale.
200 browser hours → 200 browser_hour → Starter $3.00/mo = $0.02 per browser hour
Every AgentQL plan
| Plan | Monthly | If billed yearly | Prepaid pack | Included | Simultaneous sessions |
|---|---|---|---|---|---|
| Starter your job | $0.00 | n/a | n/a | 50 request | 5 |
| Professional | $99.00 | n/a | n/a | 10,000 request | 100 |
What consumes a unit
This is where a bill actually grows. One operation is not one billed unit, and the multiplier is published here rather than on the pricing page.
| Operation | Costs | Per | Source |
|---|---|---|---|
| request.ai_extraction | 1 request | request | source |
| browser.session_time | 0.12 usd | browser_hour | source |
What AgentQL does not publish
- Failure billing is not published.
- The pricing page does not explain whether one extraction call and its remote-browser runtime are always additive or only additive for browser-backed SDK use.
What AgentQL is
AgentQL is a web extraction and automation product built around a semantic query language. Developers can use its REST API to retrieve structured data or use its Python and JavaScript SDKs with Playwright for extraction and browser interaction. A buyer comparing it with Browse AI, Diffbot, and Firecrawl should focus on that query layer: AgentQL is aimed at developers who want to describe the data or page element they need while retaining code-level control. It is not primarily a visual robot recorder, a web-scale knowledge graph, or a crawler-first content API. Sources: AgentQL product page, AgentQL documentation, Browse AI pricing, Diffbot pricing, and Firecrawl pricing, retrieved 2026-07-26.
How billing works
The basic extraction unit is one AgentQL API call. A plan includes a call allowance, and calls above that allowance use the plan’s per-call rate. Remote browser time is a second meter with its own included allowance and an hourly overage rate. The two meters matter separately because the REST extraction path and a browser-backed SDK workflow do not necessarily consume the same resources. The common buying mistake is to divide the subscription by the included API calls and treat that as the whole unit cost, while ignoring remote browser time in workflows that actually use it. Source: AgentQL product and pricing page, retrieved 2026-07-26.
What AgentQL does not publish
- Whether failed, blocked, or otherwise unusable API calls are billed. Source reviewed: AgentQL product and pricing page, retrieved 2026-07-26.
- A rule explaining exactly when an extraction call and remote browser runtime are additive, rather than alternative integration paths. Sources reviewed: AgentQL product page and AgentQL documentation, retrieved 2026-07-26.
- A self-serve enterprise rate for managed datasets, dedicated cloud environments, or on-premises deployment. Source reviewed: AgentQL product and pricing page, retrieved 2026-07-26.
- An annual-billing option or rollover rule for unused monthly allowances on the public pricing section. Source reviewed: AgentQL product and pricing page, retrieved 2026-07-26.
Who it suits
AgentQL suits development teams that want semantic extraction inside Playwright or a direct REST workflow and can estimate API calls and browser hours separately. It is less suitable for a buyer who needs one success-only unit covering every retry and browser resource, or for a nontechnical operations team looking for a visual recorder as the main interface. The reason is not product quality; it is that the public commercial model asks the buyer to map one setup to two usage meters, while important failure-billing detail remains unpublished. Sources: AgentQL documentation and AgentQL product and pricing page, retrieved 2026-07-26.
Answer
AgentQL's dated 2026-07-25 price snapshot gives $3.00 for 200 browser hours on its Starter offer. This older figure is excluded from rankings; check the vendor before buying.
Recorded 2026-07-25 from the vendor's own pricing page.A vendor page moved or could not be matched closely enough. Treat this as an older snapshot and check the vendor before buying.