Polling a search endpoint for "anything newer than the last thing I saw" has two failure modes. Statements do not enter the API in the order their articles were published: one can arrive after you already moved your "newer than" mark past it, and you never see it. And if you widen the window to be safe, you see the same statements again and again.
/v2/feed removes both problems. It returns statements in the order they entered the API and hands you a cursor
that marks your position in that order. Pass the cursor back and you get exactly what arrived since: nothing missed,
nothing repeated.
The first call
Without a cursor the feed starts 24 hours back. Filters are the same as on /v2/statements; here, only important
statements:
You need a RapidAPI key to run the examples: get one here.
curl -G https://finance-pulse.p.rapidapi.com/v2/feed \
--data-urlencode "importance_min=high" --data-urlencode "limit=5" \
-H "X-RapidAPI-Key: YOUR_RAPIDAPI_KEY" -H "X-RapidAPI-Host: finance-pulse.p.rapidapi.com"
{
"data": [
{
"statement": "ADNOC signed a multi-year LNG deal with Gulf Group to ship 2 million tonnes of gas starting in 2027",
"published_at": "2026-10-12T09:02:19Z",
"indexed_at": "2026-10-12T10:02:20.151038Z",
"source_domain": "offshore-energy.biz",
"importance": "high",
"topic_name": "ADNOC signs 2 Mt/y LNG deal with Gulf Group"
},
{
"statement": "Oil prices fall",
"published_at": "2026-10-12T06:00:57Z",
"indexed_at": "2026-10-12T10:02:20.151038Z",
"source_domain": "cnbc.com",
"importance": "high",
"topic_name": "Oil prices fall"
},
… 3 more
],
"next_cursor": "WzM0Nzc2OSwiZTNmYWQ2MWEtMWVmNS01NDgxLTgzNGYtZWJiZmU0NDE1ODY1Il0"
}
Statements come oldest first by indexed_at, the moment they entered the API. Compare the two timestamps on each:
published_at is usually the earlier of the two, by an amount that varies. A loop keyed on publication time would have to guess how
long to wait for late articles; the feed simply delivers each statement when it arrives.
The loop
Store next_cursor after every call and pass it on the next one. Persisting it to disk means a restart resumes where
it stopped:
import json, pathlib, time, requests
API = "https://finance-pulse.p.rapidapi.com"
HEADERS = {"X-RapidAPI-Key": "YOUR_RAPIDAPI_KEY", "X-RapidAPI-Host": "finance-pulse.p.rapidapi.com"}
STATE, LOG = pathlib.Path("feed.cursor"), pathlib.Path("statements.jsonl")
params = {"sector": "Energy", "importance_min": "high", "limit": 100}
while True:
if STATE.exists():
params["cursor"] = STATE.read_text()
page = requests.get(f"{API}/v2/feed", headers=HEADERS, params=params, timeout=30).json()
with LOG.open("a") as log:
for s in page["data"]:
log.write(json.dumps(s) + "\n")
print("ALERT", s["published_at"][:16], s["statement"], f"({s['source_domain']})")
STATE.write_text(page["next_cursor"])
if not page["more"]:
time.sleep(2700) # up to date: wait before the next poll. Each poll is one request: every 45 min
# is ~960 a month, inside the free plan; a paid plan's quota allows shorter pauses
import { readFileSync, writeFileSync, appendFileSync, existsSync } from "node:fs";
const API = "https://finance-pulse.p.rapidapi.com";
const HEADERS = { "X-RapidAPI-Key": "YOUR_RAPIDAPI_KEY", "X-RapidAPI-Host": "finance-pulse.p.rapidapi.com" };
const params = new URLSearchParams({ sector: "Energy", importance_min: "high", limit: "100" });
for (;;) {
if (existsSync("feed.cursor")) params.set("cursor", readFileSync("feed.cursor", "utf8"));
const page = await (await fetch(`${API}/v2/feed?${params}`, { headers: HEADERS })).json();
for (const s of page.data) {
appendFileSync("statements.jsonl", JSON.stringify(s) + "\n");
console.log("ALERT", s.published_at.slice(0, 16), s.statement, `(${s.source_domain})`);
}
writeFileSync("feed.cursor", page.next_cursor);
// up to date: wait before the next poll; each poll is one request (every 45 min is ~960 a month, inside the free plan)
if (!page.more) await new Promise((r) => setTimeout(r, 2_700_000));
}
While catching up, more is true and the loop calls again at once (a page can be short and still have more after
it); once you are current, more is false and the loop sleeps between polls. The Energy
filter's first page, as the loop prints it:
ALERT 2026-10-12T09:02 ADNOC signed a multi-year LNG deal with Gulf Group to ship 2 million tonnes of gas starting in 2027 (offshore-energy.biz)
ALERT 2026-10-12T06:00 Oil prices fall (cnbc.com)
ALERT 2026-10-12T09:50 Barclays' Mitul Kotecha expects elevated crude prices to keep pressure on the rupee (cnbctv18.com)
ALERT 2026-10-12T09:15 International Energy Agency urges member states to accelerate diesel releases from strategic oil reserves (euronews.com)
ALERT 2026-10-12T09:50 Ukraine accuses Russia of violating an energy ceasefire agreement. (reuters.com)
An empty page is normal
When nothing new matched your filters, the feed answers "data": [] and "more": false, usually with the same
cursor you sent. That is not an error: keep the cursor it returns and poll again later. An empty page with
"more": true means the feed skipped ahead without a match: call again at once.
{"data": [], "next_cursor": "WzI2MDI3OSwiNDk3YTg3…", "more": false, "snapshot": {…}}
Run this yourself: get an API key on RapidAPI ↗.