ZenRows vs. Scrapingdog vs. Bright Data: web scraping APIs compared for agent builders

Three scraping APIs, three billing models, one question for anyone building agents that need web data: which one will not surprise you on the invoice at 3 a.m.? The prices below come from the official pricing pages, October 2026.

The three models, one paragraph each

ZenRows is a credit subscription. Free with 5,000 credits/month, no card; Build $16/mo with 45K credits; Launch $58/mo with 250K; Growth $166/mo with 1.2M; Scale $458/mo with 5M. One credit buys: a standard page, 1 credit; with JavaScript rendering, 5; with premium proxies, 10; with both, 25. Failed requests are never charged. It is the only one of the three with an MCP Server and an agent toolkit in its own nav — the API was clearly rethought for agents.

Scrapingdog is also a credit subscription. Free with 100 credits; Lite $40/mo with 200K credits and 5 concurrent requests; Standard $90/mo with 1M and 50; Pro $200/mo with 3M and 100; Premium $350/mo with 6M and 150. A single subscription unlocks all 40+ APIs (Google SERP, Maps, Amazon and more), each with its own cost from 1 to 25 credits. Failed requests refund the credit; from Standard up, unused credits roll over; annual billing gives 2 months free.

Bright Data is a platform billed per record. The Web Scraper API costs $1.50 per 1,000 records on pay-as-you-go, with 5,000 records/month free (a pool shared across products); the $499/mo Scale plan includes 384K records and charges $1.30 per 1K after that. Around the scraper sit proxies, Web Unlocker, SERP API, Crawl API, a Scraping Browser billed per GB and ready-made datasets. Prepaid wallet; committed volume does not roll over.

What a credit actually buys

Same multipliers, different bundlesZenRows and Scrapingdog use the same ladder (1 / 5 / 10 / 25 credits for standard / JS-rendered / premium-proxy / both). The hard page costs 25x the easy one in both — the difference is what comes along: MCP and crawl primitives in ZenRows, 40+ Google and marketplace endpoints in Scrapingdog.
Bright Data: flat price per record$1.50 per 1,000 records regardless of whether the page was easy or hard (within the Web Scraper API). Predictable, but there is no cheap lane for simple pages — anyone scraping plain HTTP pays the same per record.
ConcurrencyScrapingdog caps by plan (5 to 150 concurrent); so does ZenRows (5 to 400); Bright Data advertises unlimited concurrency on pay-as-you-go. For a batch job your agent runs overnight, this number decides how long the run takes.
Failure policyAll three say failed requests are not charged. Verify per endpoint — and test under real load before trusting the estimated invoice.
Honest free tiersZenRows: 5,000 credits/month, always — enough to prototype an agent for real. Scrapingdog: 100 credits — a smoke test. Bright Data: 5,000 shared records/month, no card.

The agent-builder angle

Agent calling scraping as a toolZenRows' MCP Server lets the agent call scraping without you writing a client. In Scrapingdog, the ready endpoints (SERP, Maps, Amazon) already match the most common agent tasks. Bright Data has no first-party MCP — its strength is breadth and datasets when the agent needs a whole site.
Predictable cost when the agent iteratesAgents retry, re-run, re-scrape. In a credit model every hard page multiplies by 25 and a naive estimate breaks; in a flat per-record price the unit math stays legible. Budget the worst case (all pages hard), not the average.
Where the invoice changes ownerAt high volume of simple pages, Scrapingdog has the lowest effective per-1K (from ~$0.045/1K on Premium). For structured records, Bright Data's $1.50/1K is competitive. For cheap entry with MCP, ZenRows wins.

One worked example

An agent scraping 100K pages a month, half of them needing JavaScript rendering. ZenRows: 50K pages at 1 credit + 50K at 5 = 300K credits — the Launch plan ($58/mo) covers it. Scrapingdog: same arithmetic, 300K credits — Standard ($90/mo) covers it. Bright Data: 100K records at $1.50/1K = $150 on pay-as-you-go. Three different invoices for the same volume — and none of them is "wrong".

Who is this for

Choose ZenRows ifyou build agents in a small team, want MCP and a modern API, and need a free tier that sustains a real prototype.
Choose Scrapingdog ifyour tasks are Google/marketplace-shaped (SERP, Maps, Amazon), you want 40+ APIs in one subscription and credit rollover.
Choose Bright Data ifyou need breadth — proxies, an unlocker, whole-site datasets, enterprise compliance — and prefer paying per delivered record.

The short version

For agent builders the decision is rarely "which is better": it is which billing model you can explain to your client when the overnight run goes long. Multiplied credits (ZenRows, Scrapingdog) are cheap on easy pages and expensive on hard ones; flat per-record (Bright Data) is constant. Official pages to check the numbers: ZenRows pricing, Scrapingdog pricing and Bright Data Web Scraper API.

Where I come in

Scraping is the easy half of the pipeline. The part that breaks is proving the agent delivered the right data — schema, evidence, a gate that fails closed. That is what this store sells: the complete pack of agent-delivery skills.

Want a human to wire the guards into your project? Guided setup — or the free kickoff call first.

← all skills

Get the next article

Sent when I publish, together with the free skill used here. No spam, one-click unsubscribe.

I am also here: Portfolio · LifeLog (personal blog) · X

Payment processed by Stripe — no card data touches this server. Instant delivery once confirmed. Prices in BRL (Brazilian real) — you are charged in BRL. Terms, delivery and refund.