Solver API or headless browser?

There are two ways to get a request past a bot manager. Run a real browser and let it answer the challenge, or keep your requests as plain HTTP and hand the challenge to an API. Each is the right choice for some jobs.

How each one works

Headless browser

  • Puppeteer, Playwright or Selenium drives a real Chrome
  • The page’s own scripts run and answer the challenge
  • Stealth patches hide the signs of automation
  • Every page costs a render: scripts, images and CPU time

Solver API

  • Your code makes ordinary HTTPS requests
  • When a challenge appears, one API call returns the cookies or the payload that answers it
  • The session carries on from your own client
  • Pages are fetched as raw responses, with nothing rendered

Side by side

Solver APIHeadless browser
What you runAn HTTP clientA Chrome instance per concurrent session, and the servers under it
Cost followsSuccessful solvesMachines, memory and bandwidth, whether or not pages pass
Bandwidth per pageThe response you asked forThe whole page: scripts, styles, images, trackers
When the protection updatesThe provider patches the solver. Your code is unchangedYou update stealth patches and redeploy, often under time pressure
ConcurrencyAs many requests as your rate limit allowsBounded by CPU and RAM per browser
Interactive flowsOnly what can be expressed as requestsAnything a user can do
Protections coveredOnly the ones the provider solvesWhatever the browser manages to pass

Working out the cost

The two approaches are billed on different things, so compare them per session rather than per page. A cleared session usually serves many requests, which means the number that matters for an API is how often the site challenges you again, not how many pages you fetch.

As a worked example with notabot’s list price of €2.50 per 1,000 solves: a job that needs 20,000 solves a month costs €50 in solves, before any level discount, plus proxies and one small server for the HTTP client. For the browser approach, price the machines needed to hold your peak number of concurrent Chrome instances for the same month, add the extra proxy bandwidth of full page loads, and add the engineering time spent each time a protection changes. Use your own numbers: the answer differs a lot between a 500-page job and a 5-million-page one.

When to use which

Use a solver API when

  • The data is available from HTML or JSON responses
  • Volume is high, or needs to scale quickly
  • The protection in front of the site is one the provider solves
  • You do not want to maintain stealth patches

Use a browser when

  • The flow needs real interaction or client-side rendering you cannot reproduce
  • The site uses a protection no API covers yet
  • Volume is small and simplicity matters more than cost
  • You are exploring a site and do not know its requests yet

See what notabot solves on the coverage list, or read the Akamai SBSD walkthrough for what the API approach looks like in code.

Questions

Is a solver API always cheaper than a headless browser?

Not always. For a few hundred pages a day a single browser on a small server is cheap and simple. The API wins as volume grows, because its cost follows successful solves while a browser fleet’s cost follows machines.

Can I combine the two?

Yes, and it is common. Use the API to clear the protection and plain HTTP for the bulk of requests, and keep a browser for the few flows that need real rendering or interaction.

Do I still need proxies with a solver API?

Yes. Clearance is tied to an IP, so you need residential or mobile proxies either way. The API replaces the browser, not the proxy.

What does my own HTTP client need?

A TLS fingerprint that matches Chrome, the user agent that came with the solve, and a cookie jar. curl-impersonate, httpcloak and tls-client all work.

Solvers

Page updated .

Try it on your own target.

Login, copy a key, send one request. If the solve fails, it costs you nothing.

Solver API vs headless browser for anti-bot | notabot