Serper is how my agents search the web. One API key, a query, and Google's results come back as JSON in a second or two. It's very useful, and cheap: a weekly report's research runs about 100 searches, a tenth of a dollar at the smallest pack.
It's what the weekly report runs on. Every Monday, the research agents search for what changed in the tools I use and the ones challenging them: the past week only, news results, X posts and Reddit threads through site: queries. A small script in front of Serper caps each run at 150 searches and logs every query, so I can see what the agents looked for.
It's also the cheap way to give your own agent a search tool. Claude's built-in web search on the API costs $10 per 1,000 searches, plus the result tokens. Serper is $1 per 1,000 at the smallest pack, $0.30 at volume, and you choose what reaches the model.
What it isn't is an SEO tool. It gives you Google's first page, now, for one query. For rankings tracked over time, keyword volumes, backlinks and audits, I use DataForSEO. The two overlap on one thing, Google's results page, and Serper or DataForSEO? explains which to call when.
Use it when
- An agent that needs to search the web while it works: research, news of the week, fact checks, finding a page to read. Results come back in 1 to 2 seconds, so the agent doesn't wait.
- Your own agent built on the API: Claude's server-side web search costs $10 per 1,000 searches plus the result tokens; Serper is $0.30 to $1.00 per 1,000.
- Google's other result types from the same key: news, images, videos, places, maps, shopping, scholar, patents and autocomplete.
- Recent results only: tbs=qdr:w limits a search to the past week, and site: queries reach X posts or Reddit threads that Google has indexed.
Don't use it when
- Measuring SEO: rank tracking, keyword volumes, backlinks and audits. That's DataForSEO, with precise locations, deeper results and history.
- Thousands of searches that can wait a few minutes: DataForSEO's Standard queue is $0.60 per 1,000 and its balance never expires.
Pricing
Prepaid credits, one per search: 2,500 free at sign-up, no card. Then $50 for 50,000 ($1.00 per 1,000), down to $0.30 per 1,000 at 12.5 million. Credits are valid for 6 months. Only successful searches are charged.
Gotchas, with fixes
Gotcha: Prepaid credits expire 6 months after purchase.
Fix: Buy the smallest pack that covers six months of real use, not the cheapest rate per 1,000.
Gotcha: Asking for 100 results still returns 10: in my test on 26 October 2026, num set to 100 gave 10 organic results for one credit.
Fix: Page through with the page parameter, one credit per page, or use DataForSEO when you need the top 100.
Gotcha: An agent in a loop can search hundreds of times in one run.
Fix: Put the calls behind your own small script that caps searches per run and logs every query, as the weekly report does.
Instead, consider
- DataForSEO SEO data: rankings by city, the top 100, keyword volumes, backlinks and audits.
- Anthropic API (for client systems) Claude's own web search tool, when an agent's few searches aren't worth wiring an API.
Decision guides
Sources
Last reviewed by Nicolas Dao