Every search below runs live through a residential exit and comes back parsed — the same call your integration makes. Nothing here is cached.
Everything above runs through the same public API. With your API key, one HTTPS call returns the parsed SERP as JSON — no browser, no scraping.
Call the fleet directly — serpapi.exalive.ai round-robins across the
serving fleet (no gateway in the path). The tier is the port:
Ultra Fast: POST http://serpapi.exalive.ai:60005/api/serp
Standard: POST http://serpapi.exalive.ai:60006/api/serp
Header: X-API-Key: sk-your-token
Body: application/json
Instead of DNS you can fetch the list of servers that are serving right now and spread your requests across all of them. A server is listed only after it passed live searches on both ports, and is dropped as soon as it fails a check. Refresh the list every minute or two; call any listed IP directly, on either port, with your key.
GET https://serp.exalive.ai/api/hosts
{"updated_at":"2026-11-07T19:30:00Z", "refresh_seconds":60,
"ports":{"ultra":60005,"standard":60006},
"hosts":["23.111.143.170","51.195.24.13", …],
"countries":{"de":["79.127.178.226", …], "gb":[…], …}}
countries holds the servers behind each
serpapi-<cc> hostname (see Country endpoints). Use those when
you search with that gl.
resolve_urls.curl -X POST http://serpapi.exalive.ai:60005/api/serp \
-H "X-API-Key: sk-your-token" \
-H "Content-Type: application/json" \
-d '{"keyword":"best running shoes","gl":"us"}'
Top ~50 results in one call: pages fetches several result
pages and returns them merged, ranked continuously:
curl -X POST http://serpapi.exalive.ai:60005/api/serp \
-H "X-API-Key: sk-your-token" \
-H "Content-Type: application/json" \
-d '{"keyword":"best running shoes","gl":"us","pages":5}'
Standard tier (port 60006) with every link on the page and full URLs resolved:
curl -X POST http://serpapi.exalive.ai:60006/api/serp \
-H "X-API-Key: sk-your-token" \
-H "Content-Type: application/json" \
-d '{"keyword":"best web hosting","gl":"us","all_links":true,"resolve_urls":true}'
Mobile results: send a phone browser's User-Agent in
user_agent and Google returns its mobile SERP, which ranks and lays
out results differently from desktop (more People Also Ask, a different #1 more
often). Desktop user agents return the same results as the default.
curl -X POST http://serpapi.exalive.ai:60005/api/serp \
-H "X-API-Key: sk-your-token" \
-H "Content-Type: application/json" \
-d '{"keyword":"pizza near me","gl":"us",
"user_agent":"Mozilla/5.0 (iPhone; CPU iPhone OS 18_1 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/18.1 Mobile/15E148 Safari/604.1"}'
| Field | Type | Default | Description |
|---|---|---|---|
keyword | string | — | Required. The search query. |
gl | string | us | Country (geo) code, e.g. us, gb, de. The search is routed through a residential exit and a Google session in that country, so you get what a local user sees (which can include foreign domains when Google ranks them there). If the server you called has no exits for that country, it forwards the request to one that does; you don't need to do anything. gl=uk is treated as gb. The country used is returned in meta.exit_country. See Country endpoints. |
hl | string | en | Interface language. |
uule | string | — | Optional location for local results, in Google's own uule format (e.g. w+CAIQICIVQmVybGluLEJlcmxpbixHZXJtYW55 = Berlin). gl sets the country; uule narrows it to a city, which matters for local queries ("dentist", "pizza delivery"). Sent to Google only when you include it. It adds about 0.2–0.3 s. |
pages | int | 1 | Result pages to fetch, 1–10. Pages are fetched in parallel and merged into one ranked organic list (duplicates removed); ads, People Also Ask and other features come from page 1. Google shows about 8 organic results per page, so 5 pages ≈ 40 results. Each page counts as one request. |
start | int | 0 | Result offset (0, 10, 20…). With pages, the first page fetched. |
num | int | — | Ignored: Google always returns one page of results. Use pages. |
all_links | bool | false | Return every destination on the page (organic, sitelinks, videos, ads), each tagged with a type. |
resolve_urls | bool | false | Standard tier only. Follow Google's redirect wrapper to recover the full URL path instead of only the domain. |
user_agent | string | — | The User-Agent Google sees for this search. Omit it for the default device. A phone UA (Android, iPhone) returns Google's mobile results; desktop UAs return desktop results. Max 512 characters. Echoed back in meta.user_agent. Works on both tiers and with every other field. On mobile results related_searches is not returned yet. |
format | string | json | Set to html to get the raw Google HTML instead of parsed JSON. |
JSON with organic (position, title, url, snippet),
people_also_ask, related_searches, optional
all_links, and a meta block with timing. With pages,
meta.pages_ok and meta.pages report each page; a page that
fails is listed there and you still get the rest. meta.exit_country (also the
X-SERP-Exit-Country header) is the country of the exit that served the search, e.g.
DE. It's exactly what
the JSON tab above shows. Authenticate with your own key; an invalid or missing
key returns 401.
Any server answers any gl: requests for a country it has no exits for
are forwarded to one that has them. To skip that extra hop, call the country's own hostname. It
points only at servers that hold that country's exits and Google sessions, and
gl defaults to that country when you leave it out:
POST http://serpapi-de.exalive.ai:60005/api/serp # Germany
POST http://serpapi-gb.exalive.ai:60005/api/serp # United Kingdom
serpapi-ca serpapi-au serpapi-fr serpapi-es serpapi-it
serpapi-nl serpapi-ie serpapi-mx serpapi-br serpapi-in
curl -X POST http://serpapi-de.exalive.ai:60005/api/serp \
-H "X-API-Key: sk-your-token" \
-H "Content-Type: application/json" \
-d '{"keyword":"versicherung vergleich","hl":"de"}'
meta.exit_country shows where the search actually ran. Check it when the
country matters.