NOAA Tides API - Tide Predictions, High Tide & Low Tide Times

For boating, fishing and surf apps, coastal construction scheduling, and any calendar that has to know when the water moves: official NOAA tide predictions for the United States — the time and height of every high tide and low tide, for any of NOAA's 3,499 tide prediction stations, from the CO-OPS API.

Every station type works, not just the easy ones: 1,256 reference stations and 2,243 subordinate stations (measured from the official station list on 2026-08-31). Give it a station id, or just a lat,lon point and it finds the nearest station for you — 13.44,144.66 resolved to Apra Harbor, Guam at 0.54 km, with a +10 offset. Up to 365 days in one run, and past dates work.

The raw API returns local clock time with no offset (2026-09-01 04:32), which no scheduler can use. This Actor reads the same window in the station's local time and in UTC and lines the two up, so every row carries timeLocal, timeUtc and utcOffsetHours — plus minutesToNextTide and the range to it, already computed.

NOAA also rejects a request with HTTP 200 and an error body. That is reported as a rejection, never as "no tides".

• No API key, no scraping, no proxy. US government data, public domain.
• Past and future dates. Tide tables for a whole month, or just the next tides from right now.
• Both unit systems on every row — feet and metres.
• Export as JSON, CSV or Excel.

What this returns that the raw NOAA endpoint does not

The CO-OPS predictions endpoint answers with three fields per tide: a time string, a height, and H or L. Everything below is the work this Actor does on top of it.

Column	Why it matters
timeUtc, utcOffsetHours	The raw API returns local clock time with no offset (2026-09-01 04:32). A scheduler cannot use that. This Actor reads the same window in the station's local time and in UTC and lines the two up, so every tide carries a real ISO-8601 UTC timestamp and the offset that applied at that moment — daylight saving included. When the two cannot be lined up, timeUtc is null and the run says so: it never guesses an offset.
stationId, stationName, state, latitude, longitude	The raw API returns none of these with the predictions.
distanceKm, requestedPoint	Give "37.81,-122.47" and you get the nearest station plus how far away it is.
heightMeters	The raw API returns one unit system per request. Both are here (meters converted from the feet NOAA returns; within 1 cm of NOAA's own metric answer).
minutesToNextTide, rangeToNextFeet, rangeToNextMeters	How long until the water turns, and how far it moves.
isNextTide	In the default "days ahead" mode, exactly one row per station is marked as the next tide from the moment the run starts. In a fixed beginDate/endDate window it marks the first tide that is still ahead, and nothing at all for a window in the past.
timeZoneName	NOAA's own name for the station's standard time zone (EST even in July). Use utcOffsetHours for the offset that actually applied to that tide - it follows daylight saving.
stationType, referenceStationId	Whether a station is a NOAA reference station or a subordinate station, and which station it is derived from.

Input

Field	Type	Default	Meaning
locations	array	["8443970"]	NOAA station ids (8443970, TEC1603) and/or points ("42.36,-71.05"). Points are resolved to the nearest tide prediction station.
days	integer	7	Days of tides from the moment the run starts. Only tides still ahead are returned. About 4 tides per day per station.
beginDate	string	""	Start of a fixed window, "2026-09-01". The station's own local calendar days, like a printed tide table. Past dates work.
endDate	string	""	End of the fixed window. At most 365 days in one run.
datum	string	"MLLW"	MLLW, MLW, MSL, MTL, MHW, MHHW, NAVD or STND.
maxRetries	integer	4	Retries when NOAA does not answer. Bad requests are not retried.

Output

One row per predicted high or low tide:

{
  "rowType": "tide",
  "stationId": "8443970",
  "stationName": "Boston",
  "state": "MA",
  "latitude": 42.35389,
  "longitude": -71.05028,
  "stationType": "reference",
  "tideType": "Mixed",
  "tide": "high",
  "tideCode": "H",
  "timeLocal": "2026-08-31 14:05",
  "dateLocal": "2026-08-31",
  "timeUtc": "2026-08-31T18:05:00.000Z",
  "utcOffsetHours": -4,
  "heightFeet": 10.43,
  "heightMeters": 3.179,
  "datum": "MLLW",
  "minutesToNextTide": 375,
  "rangeToNextFeet": 10.439,
  "rangeToNextMeters": 3.182,
  "isNextTide": true,
  "source": "NOAA CO-OPS Tides & Currents (official, public domain)"
}

What it does when something is wrong

A tide feed that quietly answers "no tides" is worse than one that fails, so this Actor never does that.

• A station id that does not exist, a station with no predictions (the Great Lakes have none), or a datum the station does not publish → one row saying which input failed and why, and that row is not charged.
• NOAA rejecting a request with HTTP 200 and an error body — which is what happens when you ask for a datum a station does not carry — is treated as a failure, not as a day without tides.
• If no location at all could be read, the run fails instead of finishing green with an empty dataset.
• If the local and UTC copies of the predictions cannot be lined up, the rows still arrive with local time and heights, timeUtc is null, and the run logs it and records RUN_UTC_UNMATCHED.
• A run that hits the maximum total charge you set says so in the log and in RUN_BUDGET_STOP.

Measured

Run on 2026-08-31 with 5 inputs — 2 station ids, 2 points, 1 station id that does not exist: 35 tide rows + 1 error row in 7.9 seconds, 0 rows with a missing UTC timestamp. 13.44,144.66 resolved to Apra Harbor, Guam at 0.54 km, with a +10 offset. 42.36,-71.05 resolved to Boston at 0.68 km — the same station as the id 8443970 in the same run, so it was read once and charged once, not twice. A 365-day window was checked at Boston, Troy, Guam and Hawaii: both daylight saving changes were picked up in the right places, and no timestamp went backwards.

The build is covered by 70 automated checks, and 65 deliberate defects were injected one at a time to confirm those checks actually catch them.

Billing

Pay per event: one charge per tide row delivered. Rows that report a problem, and the row that says "no tides in this window", are not charged. Setting a maximum total charge on the run stops delivery rather than handing rows over unpaid.

Support

Questions and bug reports: use the Issues tab on this Actor. Include the input you ran and the run id.

Data source: NOAA CO-OPS Tides & Currents — official US government data, public domain. This Actor is not affiliated with NOAA.