186,148 river and water API endpoints that answered a keyless GET with a data body, tested 10–14 Nov 2026. For agent and app builders.
Compiled by AgentPopulation with AP Desk · @agentpopulation
What's in it · For AI agents · Quick start · Test it yourself · Files · Licence
For the latest tested endpoints, search agentpopulation.com/apis. This pack is a snapshot from 14 Nov 2026.
This repository holds a 1,000-row sample and the services list; the full list of 186,148 endpoints is one download from the latest release.
Endpoints for building a river, flood or tide agent or app. River gauges, flood stations, tide and current predictions, ocean datasets and groundwater wells, as URLs that answered with no key on their test date.
AgentPopulation has tested over 1 million API endpoints: 1,411,945 tested, 1,033,784 passed a data call (snapshot 14 Nov 2026). We are publishing them pack by pack, and more packs are coming. This is the river and water pack: 186,148 endpoints from 8,401 services. Packs overlap, so an endpoint can appear in more than one. See All packs.
Selection rule: rows where tags has rivers OR tags has floods OR tags has tides OR tags has oceans OR tags has groundwater OR tags has water.
What is an API endpoint? An API endpoint is one web address that a program can ask for data, for example the latest reading at one weather station. A single service, such as a national weather agency, can have thousands of endpoints: one per station, city or dataset. This pack counts endpoints, not separate services.
What does "passed a data call" mean? On the date shown in each row, AgentPopulation sent one plain request to that address, with no password or key, and got a non-empty data body back (JSON, CSV, XML or a file): not an error page and not an empty result. Some bodies describe a dataset rather than hold its values; see Known limits. It shows the address answered with data on that day. It is not a promise that it answers today, and licences and usage limits were not reviewed.
Four ways to use this pack. Each bucket gives a count, a filter you can apply to the CSV, one example that passed, and a list of recommended starting points with the rule that picked them. Buckets overlap; a row can fit more than one.
Examples, recommended lists and the preview are drawn from rows that match this pack's focus rule, so a row that a tag put in this pack but that is about something else is not shown as a starting point. Counts cover the whole bucket.
The focus rule
Focus rule: category in ("Hydrology & Oceans", "Hazards", "Maritime") AND NOT title matches /\(\)$/i AND title matches /\b(hub.eau|usgs|noaa|nappes?|river|water|level|gauge|flow|discharge|tide|tidal|ocean|sea|wave|groundwater|well|flood|rain|hydro|lake|reservoir|stream|buoy|current|piezo|erddap|niveau|débit|pegel|wasser|coast|marine)/i
38,658 endpoints from 3,462 services · Wrap a gauge URL as a get_river_level tool: pass the site id, return the latest level and time.
13,333 of these are ArcGIS map layers (river, water level, hydrology) rather than gauge readings.
Start with: tags has agent-ready AND tags has rivers AND subcategory matches /river|water level|gauge|discharge|hydrology|waterway/i AND NOT url contains "/erddap/info/" AND NOT title matches /\(\)$/i
Example: USGS real-time: St. John River At Ninemile Bridge… · USGS Water Services · passed 2026-11-14
Recommended starting points: 8 services, picked by the stated criteria inside
Recommended starting points. Selected by stated criteria, not a ranking of quality: Services in this pack with the most endpoints that match this bucket and are tagged agent-ready (no key sent in the test call, JSON body, passed a data call). Ties go to the service with more country codes, then by name. Only services with at least one bucket row that matches the focus rule are listed, and the sample comes from those rows; the counts cover the whole bucket. The sample URL is the first row by URL sort after preferring, in this order: rows that match the bucket's sample preference when it has one, rows that return values rather than describe a dataset or a service (not ERDDAP /info/ pages, GTFS validation reports, untitled rows or service root addresses), and agent-ready rows. Sample preference in this bucket: subcategory matches /gauge/i. Services named "ArcGIS organization" hold map layers from many owners and are added only when fewer than 5 other services match.
| Service (host) | Endpoints in bucket | In pack | Country codes in bucket | Formats in bucket | Sample URL that passed |
|---|---|---|---|---|---|
| USGS Water Services (waterservices.usgs.gov) | 11,626 | 16,342 | GU, US | json | USGS real-time: St. John River At Ninem… (2026-11-14) |
| UK Environment Agency data (environment.data.gov.uk) | 7,394 | 9,152 | GB | json | EA flood monitoring: Charlton (Thames T… (2026-11-14) |
| HubEau (French water data) (hubeau.eaufrance.fr) | 4,116 | 7,289 | FR | json | Hub'Eau real-time river level: La Grand… (2026-11-14) |
| flood-api.open-meteo.com | 835 | 835 | 186 codes | json | Open-Meteo/GloFAS river discharge near … (2026-11-14) |
| www.pegelonline.wsv.de | 787 | 788 | DE | json | PEGELONLINE water level: VERSEN TRENNSP… (2026-11-14) |
| NOAA CO-OPS tides and currents (api.tidesandcurrents.noaa.gov) | 235 | 6,569 | US | json | NOAA Water level: Nawiliwili, HI (2026-11-14) |
| hsi-sh.de | 158 | 1,125 | DE | json | Water level level Westerland W6 - North… (2026-11-14) |
| OpenDataSoft public data network (data.opendatasoft.com) | 62 | 950 | 4 codes | json | Aqueduct country and river basin rankin… (2026-11-14) |
26,320 endpoints from 3,695 services · Show flood warnings on a map and tide times on a card.
Start with: tags has floods OR tags has tides
Example: NOAA Water level: Nawiliwili, HI · NOAA CO-OPS tides and currents · passed 2026-11-14
Recommended starting points: 8 services, picked by the stated criteria inside
Recommended starting points. Selected by stated criteria, not a ranking of quality: Services in this pack with the most endpoints that match this bucket and passed a data call. Ties go to the service with more country codes, then by name. Only services with at least one bucket row that matches the focus rule are listed, and the sample comes from those rows; the counts cover the whole bucket. The sample URL is the first row by URL sort after preferring, in this order: rows that match the bucket's sample preference when it has one, rows that return values rather than describe a dataset or a service (not ERDDAP /info/ pages, GTFS validation reports, untitled rows or service root addresses), and agent-ready rows. Services named "ArcGIS organization" hold map layers from many owners and are added only when fewer than 5 other services match.
| Service (host) | Endpoints in bucket | In pack | Country codes in bucket | Formats in bucket | Sample URL that passed |
|---|---|---|---|---|---|
| NOAA CO-OPS tides and currents (api.tidesandcurrents.noaa.gov) | 6,569 | 6,569 | US | json | NOAA Water level: Nawiliwili, HI (2026-11-14) |
| UK Environment Agency data (environment.data.gov.uk) | 3,599 | 9,152 | GB | json | EA flood monitoring: Charlton (Thames T… (2026-11-14) |
| flood-api.open-meteo.com | 835 | 835 | 186 codes | json | Open-Meteo/GloFAS river discharge near … (2026-11-14) |
| data.qld.gov.au | 243 | 297 | AU | json | Sue Island (Warraber) tide gauge—predic… (2026-11-14) |
| www.ioc-sealevelmonitoring.org | 196 | 196 | 43 codes | json | IOC sea level station: Isla Hormiga,Lim… (2026-11-14) |
| geoservice.waterinfo.be | 190 | 190 | BE | xml | Flood risk map Protected areas - FLUVIA… (2026-11-14) |
| NOAA IOOS Sensors ERDDAP (erddap.sensors.ioos.us) | 147 | 28,846 | US | json | ERDDAP IOOS-Sensors: Aguchik Island, AK… (2026-11-14) |
| OpenDataSoft public data network (data.opendatasoft.com) | 134 | 950 | GB | json | City Plan 2014 — Flood overlay — Brisba… (2026-11-14) |
92,848 endpoints from 2,503 services · Load ocean, tide and groundwater series on a schedule into your own store.
67,846 of these are ERDDAP dataset descriptions (/erddap/info/); the series needs a second request.
Start with: tags has oceans OR tags has groundwater OR format is csv
Example: Hub'Eau groundwater level: Saint-Georges-sur-l'Aa… · HubEau (French water data) · passed 2026-11-14
Recommended starting points: 8 services, picked by the stated criteria inside
Recommended starting points. Selected by stated criteria, not a ranking of quality: Services in this pack with the most endpoints that match this bucket and passed a data call. Ties go to the service with more country codes, then by name. Only services with at least one bucket row that matches the focus rule are listed, and the sample comes from those rows; the counts cover the whole bucket. The sample URL is the first row by URL sort after preferring, in this order: rows that match the bucket's sample preference when it has one, rows that return values rather than describe a dataset or a service (not ERDDAP /info/ pages, GTFS validation reports, untitled rows or service root addresses), and agent-ready rows. Sample preference in this bucket: NOT url contains "/erddap/info/". Services named "ArcGIS organization" hold map layers from many owners and are added only when fewer than 5 other services match.
| Service (host) | Endpoints in bucket | In pack | Country codes in bucket | Formats in bucket | Sample URL that passed |
|---|---|---|---|---|---|
| NOAA IOOS Sensors ERDDAP (erddap.sensors.ioos.us) | 28,846 | 28,846 | US | json | ERDDAP IOOS-Sensors: 067 - San Nicolas … (2026-11-14) |
| GRIIDC ERDDAP (Gulf of Mexico) (erddap.griidc.org) | 11,688 | 11,688 | global or unknown | json | ERDDAP GRIIDC: Trajectory - F2.x836.000… (2026-11-14) |
| NOAA PFEG / UAF ERDDAP (upwell.pfeg.noaa.gov) | 8,707 | 8,707 | US | json | ERDDAP UAF: PacIOOS Weather Station: Ho… (2026-11-14) |
| HubEau (French water data) (hubeau.eaufrance.fr) | 3,174 | 7,289 | FR | json | Hub'Eau groundwater level: Saint-George… (2026-11-14) |
| USGS Water Services (waterservices.usgs.gov) | 3,173 | 16,342 | PR, US | json | USGS real-time: Barnegat B Tdl Chanl1 A… (2026-11-14) |
| coastwatch.pfeg.noaa.gov | 2,989 | 2,989 | US | json | ERDDAP CSWC: CalCOFI Farallon Institute… (2026-11-14) |
| gliders.ioos.us | 2,619 | 2,619 | US | json | ERDDAP NGDAC: Gichigami-20110629T1821 (2026-11-14) |
| waterlevel.ie | 2,033 | 2,033 | IE | csv | OPW Hydrology Reference Network (2026-11-14) |
186,148 endpoints · 8,401 services · 245 country codes · 7 formats · Run your tool calls against a fixed set of URLs across formats and record which answer; last_tested is the baseline for your own re-tests.
Start with: the whole pack
Example: USGS real-time: St. John River At Ninemile Bridge… · USGS Water Services · passed 2026-11-14
Recommended starting points: 8 services, picked by the stated criteria inside
Recommended starting points. Selected by stated criteria, not a ranking of quality: Services in this pack that returned the most different formats, then the most endpoints that passed a data call, then by name: a mix of response shapes to test parsers and tool calls. Only services with at least one bucket row that matches the focus rule are listed, and the sample comes from those rows; the counts cover the whole bucket. The sample URL is the first row by URL sort after preferring, in this order: rows that match the bucket's sample preference when it has one, rows that return values rather than describe a dataset or a service (not ERDDAP /info/ pages, GTFS validation reports, untitled rows or service root addresses), and agent-ready rows. Services named "ArcGIS organization" hold map layers from many owners and are added only when fewer than 5 other services match.
| Service (host) | Endpoints in bucket | In pack | Country codes in bucket | Formats in bucket | Sample URL that passed |
|---|---|---|---|---|---|
| www.odwb.be | 48 | 48 | BE | xml, json, txt, csv, geojson | Hydrogeological map (2026-11-14) |
| BSH geospatial data (Germany) (gdi.bsh.de) | 178 | 178 | DE | xml, csv, txt, json | Marine regions (2026-11-14) |
| vogis.cnv.at | 90 | 90 | AT | xml, csv, json, geojson | 5m depth lines Lake Constance east of t… (2026-11-14) |
| data.bs.ch | 60 | 60 | CH | json, csv, txt, xml | Environmental analysis of groundwater (2026-11-14) |
| geo.rijkswaterstaat.nl | 36 | 36 | NL | xml, csv, geojson, zip | Indication of floodplain sub-area (2026-11-14) |
| data.brno.cz | 35 | 35 | CZ | json, csv, zip, geojson | Flood zones - 5 year floods (2026-11-14) |
| data.tg.ch | 34 | 34 | CH | json, csv, xml, txt | hydrologische-messnetze (2026-11-12) |
| geo.sv.rostock.de | 28 | 28 | DE | xml, csv, json, geojson | Highly imminent groundwater in the Hans… (2026-11-14) |
The same lists are in recommended.json, ready for an agent to load.
- What is the latest water level at a NOAA tide station? Try NOAA Water level: Nawiliwili, HI (NOAA CO-OPS tides and currents, passed 2026-11-14).
- How has a US river gauge moved over the last six hours? Try USGS real-time: St. John River At Ninemile Bridge… (USGS Water Services, passed 2026-11-14).
- What river discharge is forecast near a given point? Try Open-Meteo/GloFAS river discharge near Fuvahmulah… (flood-api.open-meteo.com, passed 2026-11-14).
Note
On the date in last_tested (IST), one GET to the URL, sent with no key, returned HTTP 2xx and a non-empty data body. HTML pages and empty results did not pass. The format column records what came back. This is a test result for that date, not a promise that the URL answers today. Licences, terms, rate limits, accuracy and freshness were not reviewed. Endpoints are URLs, not separate APIs; many share one service. This pack was tested 10–14 Nov 2026; snapshot 14 Nov 2026.
Start from keyless JSON endpoints (agent-ready). The code below loads recommended.json and calls the first recommended service in the Agent tools bucket, with a timeout (node examples/agent-start.mjs --recommended). With no argument, node examples/agent-start.mjs calls the most recently tested agent-ready row of the sample; pass a .csv or .csv.gz path to use that file instead. See examples/agent-start.mjs.
// Node 18+. Loads recommended.json (the local copy, or the raw file on GitHub) and calls one keyless JSON endpoint.
import fs from 'node:fs';
const RAW = 'https://raw.githubusercontent.com/agentpopulation/water-api-endpoints/main/recommended.json';
const rec = fs.existsSync('recommended.json')
? JSON.parse(fs.readFileSync('recommended.json', 'utf8'))
: await (await fetch(RAW)).json();
const pick = rec.buckets.find((b) => b.id === 'agent').services[0];
try {
const res = await fetch(pick.sample_url, {
headers: { 'User-Agent': 'agent-example/1.0' },
signal: AbortSignal.timeout(15000),
});
console.log(pick.service, res.status, res.headers.get('content-type'));
console.log((await res.text()).slice(0, 300));
} catch (err) {
console.log(`did not answer now; last passed ${pick.sample_last_tested} (${err.message})`);
}How an agent picks an endpoint:
- Filter to
agent-readyand your country or topic. - Prefer the newest
last_testedand the larger services. - Call with a timeout and check the body before you use it.
The examples run against the sample in data/ out of the box (1,000 rows). Each takes an optional path to a .csv or .csv.gz file instead.
python examples/quickstart.py
node examples/quickstart.mjs
node examples/agent-start.mjsFull list: download Water_API_Endpoints_AgentPopulation_Snapshot_2026-11-14.csv.gz (186,148 rows, 4.2 MB) from releases/latest, then gunzip it and pass its path. The scripts also read the .csv.gz as it is.
gunzip Water_API_Endpoints_AgentPopulation_Snapshot_2026-11-14.csv.gz
python examples/quickstart.py Water_API_Endpoints_AgentPopulation_Snapshot_2026-11-14.csv
node examples/quickstart.mjs Water_API_Endpoints_AgentPopulation_Snapshot_2026-11-14.csvQuote the URL: some contain commas.
curl -sL "https://raw.githubusercontent.com/agentpopulation/water-api-endpoints/main/data/Water_API_Endpoints_AgentPopulation_Sample.csv" | head -n 4
curl -s --max-time 15 "https://waterservices.usgs.gov/nwis/iv/?format=json&sites=01010000&period=PT6H" | head -c 600Standard library only: examples/quickstart.py reads the sample, or the path you pass, keeps agent-ready rows and calls the most recently tested one.
The code
# Python 3, standard library only. Run from a clone of this repository:
# python examples/quickstart.py [path]
# With no path it reads the sample in data/. Pass the full list from the latest release (.csv or .csv.gz) instead.
# Keeps agent-ready rows and calls the most recently tested one.
import csv
import gzip
import os
import sys
import urllib.request
HERE = os.path.dirname(os.path.abspath(__file__))
path = sys.argv[1] if len(sys.argv) > 1 else os.path.join(HERE, "..", "data", "Water_API_Endpoints_AgentPopulation_Sample.csv")
opener = gzip.open if path.endswith(".gz") else open
with opener(path, "rt", newline="", encoding="utf-8") as f:
rows = list(csv.DictReader(f))
agent = [r for r in rows if "agent-ready" in r["tags"].split("|")]
agent.sort(key=lambda r: r["last_tested"], reverse=True)
print(len(rows), "rows,", len(agent), "agent-ready")
if not agent:
sys.exit(0)
row = agent[0]
req = urllib.request.Request(row["url"], headers={"User-Agent": "agent-example/1.0"})
try:
with urllib.request.urlopen(req, timeout=15) as res:
body = res.read(1_000_000).decode("utf-8", "replace")
print(res.status, res.headers.get("Content-Type"))
print(body[:300])
except Exception as err:
print(f"did not answer now; last passed {row['last_tested']} ({err})")Reads the sample as JSON because the CSV has quoted commas.
The code
// Node 18+. Reads the sample as JSON (the CSV has quoted commas), local file first.
import fs from 'node:fs';
const FILE = 'data/Water_API_Endpoints_AgentPopulation_Sample.json';
const rows = fs.existsSync(FILE)
? JSON.parse(fs.readFileSync(FILE, 'utf8'))
: await (await fetch('https://raw.githubusercontent.com/agentpopulation/water-api-endpoints/main/data/Water_API_Endpoints_AgentPopulation_Sample.json')).json();
const row = rows.find((r) => r.tags.includes('agent-ready'));
const res = await fetch(row.url, { headers: { 'User-Agent': 'agent-example/1.0' }, signal: AbortSignal.timeout(15000) });
console.log(res.status, (await res.text()).slice(0, 300));examples/agent-tools.json holds two tool definitions in the shape most tool-calling APIs accept (name, description, JSON Schema); rename parameters if yours uses another key. examples/quickstart.mjs implements them:
find_endpointssearches the list by words, country code and format.fetch_endpointcalls one URL, refuses any URL that is not in the list and does not follow redirects. The allowlist is built from the CSV, so a URL injected into a prompt is refused.
The same two tools can sit behind an MCP server.
Re-test endpoints from the sample with the same rule: HTTP 2xx and a non-empty body that is not an HTML page. Node 18+, no packages.
node tools/test-endpoints.mjsIt tests the first 50 URLs and prints one line: tested, passed, did not pass. Options: a .csv or .csv.gz path (the full list from the release), --limit 50, --timeout-ms 10000, --out results.csv. The results file holds url, status, passed, content_type, ms and checked_at; a status of timeout or error:<code> means no HTTP answer came back. It is polite: GET only, no key sent, at most 4 requests in flight, one at a time per listed host, at least 1 second between requests to one listed host, and reading stops once 256 kB of a body is kept. Redirects are followed and may lead to another host. results.csv is in .gitignore.
One row from each of the five largest services among rows that match the focus rule.
| Title | Service | Country | Format | Tested | Link |
|---|---|---|---|---|---|
| ERDDAP IOOS-Sensors: 067 - San Nicolas Island, CA (46219) | NOAA IOOS Sensors ERDDAP | US | json | 2026-11-14 | erddap.sensors.ioos.us |
| USGS real-time: St. John River At Ninemile Bridge, Maine | USGS Water Services | US | json | 2026-11-14 | waterservices.usgs.gov |
| ERDDAP GRIIDC: Trajectory - F2.x836.000.0005 - Backscatter_… | GRIIDC ERDDAP (Gulf of Mexico) | global or unknown | json | 2026-11-14 | erddap.griidc.org |
| ERDDAP UAF: PacIOOS Weather Station: Honolulu Harbor Entran… | NOAA PFEG / UAF ERDDAP | US | json | 2026-11-14 | upwell.pfeg.noaa.gov |
| EA flood monitoring: Charlton (Thames Tideway) | UK Environment Agency data | GB | json | 2026-11-14 | environment.data.gov.uk |
Browse the 1,000-row sample: Sample CSV · Sample JSON.
Country codes: 245
| Code | Country | Endpoints | |
|---|---|---|---|
| US | United States | 109,513 | ████████████████████ |
| GB | United Kingdom | 13,742 | ██▌ |
| FR | France | 7,655 | █▍ |
| NZ | New Zealand | 6,058 | █▏ |
| DE | Germany | 3,998 | ▊ |
| CA | Canada | 3,824 | ▊ |
| IE | Ireland | 3,125 | ▋ |
| BE | Belgium | 2,003 | ▍ |
| EU | European Union | 1,392 | ▎ |
| AU | Australia | 1,222 | ▎ |
| other 235 codes | 6,226 | ||
| global or unknown | 27,390 | █████ |
Categories: top 8
| Category | Endpoints | |
|---|---|---|
| Hydrology & Oceans | 170,176 | ████████████████████ |
| Weather & Climate | 6,592 | ▊ |
| Hazards | 4,129 | ▌ |
| Maritime | 2,297 | ▎ |
| Government & Open Data | 1,716 | ▎ |
| Environment | 704 | ▏ |
| Aviation | 127 | ▏ |
| Navigation & Geo/Maps | 106 | ▏ |
Formats: json 168,497 · xml 9,573 · csv 4,076 · geojson 2,410 · txt 947 · zip 636 · gtfs 9
Largest services: top 25 of 8,401
| Service | Host | Endpoints | Keyless JSON | |
|---|---|---|---|---|
| NOAA IOOS Sensors ERDDAP | erddap.sensors.ioos.us | 28,846 | 28,846 | ████████████████████ |
| USGS Water Services | waterservices.usgs.gov | 16,342 | 16,342 | ███████████▍ |
| GRIIDC ERDDAP (Gulf of Mexico) | erddap.griidc.org | 11,688 | 11,688 | ████████▏ |
| UK Environment Agency data | environment.data.gov.uk | 9,152 | 9,152 | ██████▍ |
| NOAA PFEG / UAF ERDDAP | upwell.pfeg.noaa.gov | 8,707 | 8,707 | ██████ |
| HubEau (French water data) | hubeau.eaufrance.fr | 7,289 | 7,289 | █████ |
| NOAA CO-OPS tides and currents | api.tidesandcurrents.noaa.gov | 6,569 | 6,569 | ████▌ |
| coastwatch.pfeg.noaa.gov | coastwatch.pfeg.noaa.gov | 2,989 | 2,989 | ██▏ |
| US National Weather Service (api.weather.gov) | api.weather.gov | 2,735 | 2,733 | █▉ |
| gliders.ioos.us | gliders.ioos.us | 2,619 | 2,619 | █▉ |
| Environment and Climate Change Canada GeoMet/MSC | api.weather.gc.ca | 2,224 | 21 | █▌ |
| timeseries.sepa.org.uk | timeseries.sepa.org.uk | 2,216 | 2,216 | █▌ |
| hilltop.gw.govt.nz | hilltop.gw.govt.nz | 2,049 | 0 | █▍ |
| waterlevel.ie | waterlevel.ie | 2,033 | 0 | █▍ |
| erddap.bco-dmo.org | erddap.bco-dmo.org | 1,918 | 1,918 | █▍ |
| DOV Flanders subsoil / water | www.dov.vlaanderen.be | 1,659 | 0 | █▏ |
| geoport.usgs.esipfed.org | geoport.usgs.esipfed.org | 1,469 | 1,469 | █ |
| erddap.observations.voiceoftheocean.org | erddap.observations.voiceoftheocean.org | 1,375 | 1,375 | █ |
| hsi-sh.de | hsi-sh.de | 1,125 | 738 | ▊ |
| coastwatch.noaa.gov | coastwatch.noaa.gov | 1,082 | 1,082 | ▊ |
| polarwatch.noaa.gov | polarwatch.noaa.gov | 1,080 | 1,080 | ▊ |
| hilltop.nrc.govt.nz | hilltop.nrc.govt.nz | 983 | 0 | ▋ |
| OpenDataSoft public data network | data.opendatasoft.com | 950 | 950 | ▋ |
| pndb.opendatasoft.com | pndb.opendatasoft.com | 871 | 871 | ▋ |
| ArcGIS Online layers (services6.arcgis.com, org BAJNi3EgCdtQ1BCG) | services6.arcgis.com | 870 | 870 | ▋ |
Rows per test date: 10–14 Nov 2026
| Date (IST) | Endpoints | |
|---|---|---|
| 2026-11-14 | 182,641 | ████████████████████ |
| 2026-11-13 | 1,542 | ▏ |
| 2026-11-12 | 1,578 | ▏ |
| 2026-11-11 | 386 | ▏ |
| 2026-11-10 | 1 | ▏ |
| File | Rows | Size | Use |
|---|---|---|---|
| data/Water_API_Services_AgentPopulation.csv | 8,401 | 513.5 kB | One row per service with its endpoint count in this pack |
| data/Water_API_Endpoints_AgentPopulation_Sample.csv | 1,000 | 385.7 kB | The sample: focus-rule rows and agent-ready rows first, spread across services |
| data/Water_API_Endpoints_AgentPopulation_Sample.json | 1,000 | 586.7 kB | The same rows as JSON, tags as an array |
| Water_API_Endpoints_AgentPopulation_Snapshot_2026-11-14.csv.gz | 186,148 | 4.2 MB | The full list, gzipped CSV: a download from the latest release, checked by SHA256SUMS.txt |
| tools/test-endpoints.mjs | Re-tests endpoints from the sample or the full list |
Sample size: 1,000 rows, which is 1,000 or one third of the pack rounded down, whichever is smaller. Rows that match the focus rule and agent-ready rows come first, spread across services; within a service, rows that return values and the newest test dates come first. Service root addresses are placed at the end of the sample.
Full list: download Water_API_Endpoints_AgentPopulation_Snapshot_2026-11-14.csv.gz (186,148 rows, 4.2 MB) from releases/latest, then gunzip it and pass its path. The scripts also read the .csv.gz as it is.
Excel users: Data > From Text/CSV, UTF-8. Files are UTF-8 without a byte-order mark and comma-separated.
Columns, types and tags
| Column | Type | Example | Meaning |
|---|---|---|---|
| url | string | https://erddap.sensors.ioos.us/erddap/i… | The address that was tested |
| title | string | ERDDAP IOOS-Sensors: 067 - San Nicolas … | Station, dataset or feed name |
| service | string | NOAA IOOS Sensors ERDDAP | The service the endpoint belongs to; joins to the services file together with host |
| host | string | erddap.sensors.ioos.us | Host name of the address |
| country | string | US | ISO 3166-1 alpha-2 code, or EU; empty when global or unknown |
| category | string | Hydrology & Oceans | Category of the endpoint |
| subcategory | string | ERDDAP dataset (NOAA IOOS Sensors ERDDA… | Finer label, for example a station or dataset type |
| format | string | json | What the test call returned: json, csv, geojson, xml, gtfs, zip, txt and so on |
| tags | string | water|oceans|no-key|json|agent-ready |
Tags separated by | |
| last_tested | date | 2026-11-14 | Date of the test call, YYYY-MM-DD (IST) |
| Tag | Rows | Meaning |
|---|---|---|
| no-key | 186,148 | No key was sent in the test call |
| water | 170,176 | |
| agent-ready | 168,497 | No key sent, JSON body, passed a data call on last_tested |
| json | 168,497 | |
| oceans | 77,263 | |
| rivers | 63,313 | |
| maps | 52,976 | |
| floods | 16,615 | |
| science | 14,237 | |
| weather | 14,236 | |
| groundwater | 12,536 | |
| environment | 11,730 | |
| tides | 10,798 | |
| xml | 9,573 | |
| government | 8,375 | |
| statistics | 7,665 | |
| open-data | 6,547 | |
| hazards | 4,232 | |
| csv | 4,076 | |
| maritime | 2,864 | |
| geojson | 2,410 | |
| aviation | 1,708 | |
| climate | 1,254 | |
| elevation | 931 | |
| finance | 832 | |
| forecast | 761 | |
| transit | 626 | |
| health | 492 | |
| energy | 407 | |
| air-quality | 363 | |
| tsunami | 272 | |
| traffic | 210 | |
| economy | 159 | |
| parking | 89 | |
| earthquakes | 82 | |
| wildfire | 45 | |
| volcanoes | 17 | |
| geocoding | 13 | |
| gtfs | 12 | |
| embeddings | 2 | |
| ai-ml | 1 |
A machine-readable copy of this dictionary is in datapackage.json.
- 58.8% of rows carry country code US; 27,390 rows have no code (global or unknown).
- 129 rows come from PxWeb or Statistics Canada services. Some statistics endpoints return metadata or a pointer on GET; the table may need a second request or a POST.
- 67,846 rows are ERDDAP /info/ pages: they describe a dataset rather than return its values, so the series needs a second request.
- 48,268 rows are map-layer queries with resultRecordCount set: the tested call asked for a few records; change the query to get the full layer.
- 2,965 rows carry limit=1: the tested call asked for one record; raise the limit or page through.
- 3,468 rows are open-data portal dataset records, which may return a dataset record with links rather than the table itself.
- 1,583 rows are files (zip 636, txt 947) rather than bodies an agent reads directly.
- 6,569 URLs carry
application=AgentPopulation; set your own application name when you call them. - Stations and feeds can go quiet after their test date; re-test a URL before you rely on it.
- Not in this pack: endpoints that need a key or a POST, and URLs that did not pass.
- Left out: 22 rows whose URL carried a credential-looking parameter or a login.
- Left out: 1,031 ArcGIS Survey123 layers (they can hold people's form answers).
- Left out: 2 camera or image URLs (webcam and camera feeds, picture and video files).
Packs overlap. URLs in this pack that are also in another pack:
- 14,676 also in Weather API endpoints
- 900 also in Transit API endpoints
- 153 also in Finance and economy API endpoints
- 284 also in India API endpoints
- 82 also in Earthquake API endpoints
See All packs for the full list.
Important
This list (URLs, titles, tags and test results) is © 2026 AgentPopulation, released under CC BY 4.0. Credit AgentPopulation and link agentpopulation.com. The data each endpoint returns belongs to its provider and stays under that provider's terms; this licence does not cover it.
Copy-ready credit:
Rivers and water API endpoints by AgentPopulation (https://agentpopulation.com), CC BY 4.0, snapshot 14 Nov 2026.
GitHub's "Cite this repository" button reads CITATION.cff.
Spotted a wrong row or a dead URL? Tell us on X: @agentpopulation.
Questions, pack requests and reports of endpoints that stopped answering go to Discussions, or to @agentpopulation on X.
If this pack saved you time, a star helps others find it.
Compiled by AgentPopulation with AP Desk · All packs · More packs to come.