DEV Community: Moritz Kaminski

30+ flight data APIs for AI agents, plus 7 things to build

Moritz Kaminski — Tue, 08 Sep 2026 14:27:11 +0000

A delayed flight is never just a delayed flight. It changes the connection, the pickup, the last train, and whether the backup is any better.

If you are already travelling, you want to know whether the aircraft has left, where it is now and when it will really arrive. If you are booking, the questions change. Is the fare unusually low? Is there an award seat? What does the cheaper itinerary add in time and emissions?

The answers exist, but not in one place. They are split across fare engines, live aircraft feeds, airport boards, government registries and aviation weather services.

TaskFuel lets your agent discover, evaluate and use more than 100 tools. Before paying, it can preview the exact request and price, giving it the context needed to choose the right tool. You do not need to manually set up a separate account or subscription for each provider.

See TaskFuel’s current flight-search and tracking tools, including live status, airport boards and cached award availability.

What flight data can an AI agent use?

It helps to split the market into four categories:

Flight operations, status and tracking: what is scheduled, what actually happened and where the aircraft is now.
Flight search, fares and trip planning: itineraries, prices, award availability and booking options.
Airports, aircraft and ownership data: reference records, registrations, operators and due diligence.
Aviation weather, safety and derived intelligence: raw observations turned into delay risk, operational guidance, emissions estimates or news.

In the following sections you’ll find a selection of tools for inspiration. Scroll to the seven workflows to see what you can build today.

Flight operations, status and tracking APIs

These APIs tell you what’s happening: actual times, delays, gates, live positions and airport disruption.

StableTravel

StableTravel is one of the broadest options here. It covers live and historical search; status by flight number or registration; positions, tracks, maps and filed routes; airport boards and schedules; delay data; METAR and TAF; fares; and cached award availability. Its published API names SerpAPI and Google Flights for cash fares, Seats.aero for award seats, and FlightAware for operations and tracking. It powers TaskFuel’s main flight-search and tracking tool.

GoFlightLabs

GoFlightLabs covers real-time, historical and future flights. An agent can search by flight number or callsign, inspect schedules and delays, look up airline routes and use airport, airline and country reference data. It also offers fare search, so the same provider can support both planning and day-of-travel checks.

x402node

x402node focuses on the operational details passengers notice: scheduled and actual times, delay, gate, terminal, baggage belt, aircraft and codeshares. It also covers airport arrival and departure boards and FAA disruption programs such as ground stops, ground-delay programs and closures.

2s

2s puts live flight status, airport boards, schedules, aviation weather, aircraft and airport lookups, and accident records behind one provider. It suits briefs that need flight, weather and safety context together.

Agent Utility

Agent Utility uses OpenSky for live aircraft states and adds FAA airport delays and combined flight briefs. Its destination packs are especially useful for travel agents because they can include airport metadata, local weather, air quality, timezone, sunrise and sunset, and exchange rates in one response.

SuVerse Labs

SuVerse Labs combines routes, live OpenSky aircraft positions, aircraft registry data and airport lookup with decoded METAR and TAF reports, pilot reports, AIRMETs and SIGMETs. It gives an operational tool the aircraft and the conditions around it in one place.

Flight search, fares and trip-planning APIs

These APIs help you plan: ticket prices, travel estimates, timing advice, bookable offers.

StableTravel

StableTravel searches one-way, round-trip and multi-city fares with filters for cabin, stops, airline and price. It returns price context and booking links, and it searches cached award availability. Its published price for Google Flights search was 2¢ when we checked on August 13, 2026. Your agent can dry-run the exact request to quote it, then pin that amount as the maximum before paying.

BRIJ Travel

BRIJ returns live airline offers with base fare, taxes, fare brand, baggage, change and refund conditions, seller, ancillary prices and emissions. In our August 2026 endpoint checks, it was the only responding listing we found that explicitly documented booking, order-status and refund operations rather than only linking to another booking site. Its upstream flight provider is not publicly named.

SerpAPI MPP

SerpAPI’s paid endpoint provides direct Google Flights search per call. It is a straightforward option when an agent needs Google Flights results and does not need a broader travel platform around them.

GoFlightLabs

GoFlightLabs searches fares by route and date alongside its schedules, status and delay data. Its own commercial API is the stated upstream; the suppliers behind that API are not publicly disclosed.

Auor

Auor returns flight offers from Amadeus. For an agent that already wants Amadeus-shaped inventory but prefers a paid call over setting up a direct commercial account, this is the clearest match in the catalog.

ForgeMesh Fare Intelligence

ForgeMesh combines current fare observations with historical medians and a buy-now-or-wait recommendation, then supplies affiliate booking links. It refers to unnamed travel-search partners rather than disclosing a single upstream.

TransitPulse

TransitPulse handles the part fare engines usually ignore: getting to the airport. It evaluates public-transport routes, timing and feasibility from a city to an airport, optionally working backwards from the flight’s departure time.

Airport, aircraft and ownership-data APIs

Flight movement data tells you where an aircraft is. Registry and reference data tell you what it is, who registered it and whether the records raise questions.

ForgeMesh Labs

ForgeMesh resolves IATA and ICAO codes across more than 29,000 airports, returning fields such as airport name, city, country, coordinates, elevation and timezone. It bridges the place name a user types and the airport code the next API expects.

GovParse

GovParse searches the FAA registry by owner, N-number, manufacturer, model, aircraft or registrant type, location, status and registration date. It returns aircraft details and entity-resolved registrant data for research and compliance work.

Other useful reference sources

GoFlightLabs carries airport, airline and aircraft reference data. 2s adds airport and aircraft lookup, while SuVerse combines aircraft registry records with live positions.

Aviation weather, safety and derived-intelligence APIs

Raw weather reports are useful to pilots and dispatchers. Most travellers need the next layer: Will this create a delay? Is the alternate airport likely to be better? Does the itinerary carry a larger climate cost?

x402 Factory

x402 Factory turns weather into a 48-hour airport delay-risk forecast. It considers crosswinds, ceiling, visibility and convective weather, which is closer to the question an operations agent needs to answer than a raw forecast alone.

Runway Weather Oracle

Runway Weather Oracle is built for weather prediction markets, not passenger disruption. It provides airport-station temperature nowcasts, daily-high forecasts and probabilities of finishing above or below a temperature threshold; that airport signal can still feed specialised models.

MyCO2 Suite

MyCO2 Suite estimates flight CO₂e from airport codes or coordinates, cabin class and passenger count, with an optional radiative-forcing adjustment. That makes it possible to compare emissions alongside price and duration instead of calculating them after the itinerary is chosen.

Weather sources already sitting beside the flight data

2s, StableTravel and SuVerse all expose some mix of METAR, TAF or SIGMET data. TaskFuel also carries OpenWeather for local conditions and air quality. The better answer is rarely “here is the METAR.” It is “your arrival is late, heavy rain is expected and the backup route looks risky too.”

Seven flight data workflows, and what TaskFuel supports today

A provider list only matters when the sources are combined around a decision. Here are seven places to start.

1. A live flight-status and disruption assistant

Give the agent a flight number and travel date. It checks the current status, actual departure, live position and estimated arrival. If the delay crosses a threshold, it checks airport disruption, destination weather and the remaining alternatives on the route.

A practical version can start today in TaskFuel with StableTravel or 2s for the flight. x402node can add FAA disruption programs, and an OpenSky-based provider can supply a second position source. The answer should say what changed and what to do next.

A good answer might look like this:

LH400 is airborne and 41 minutes late. The estimated arrival is 15:12 local. No ground stop is active at JFK. Two later connections remain, but the shorter one now has a 34-minute transfer.

2. An airport arrivals and departures monitor

An airport monitor can watch one board, filter to a terminal, airline or destination and report only changes: a gate move, a cancellation, a cluster of delays or the last viable departure of the evening.

StableTravel and 2s cover airport activity in TaskFuel today. x402node is another option. Add weather and system-wide delay data to distinguish one late aircraft from a wider operational problem. Hotels, pickup services and anyone meeting several arrivals in a day can act on the changes immediately.

3. A fare-comparison or deal-alert agent

Instead of checking one route once, let your agent compare nearby dates, alternate airports, nonstop versus one-stop itineraries and award availability. Save the useful baseline, then repeat the search with your own schedule.

StableTravel already provides TaskFuel with fares, booking options and award seats. The other surveyed providers can widen the comparison: SerpAPI exposes Google Flights-style price context and Auor brings Amadeus offers.

The agent should alert on a decision, not every price movement: “nonstop under €150,” “business award seat under 70,000 points” or “price is 25% below the route’s usual range.”

4. An aircraft tracking and ownership-intelligence tool

Join a live transponder position to an FAA registration, aircraft profile and registrant record. That creates a useful research view for aircraft brokers, lenders, insurers, journalists and fleet analysts.

2s gives TaskFuel a starting point for aircraft lookup. A fuller version would use an OpenSky-based provider such as SuVerse or Agent Utility for movement and GovParse for registry records.

5. An aviation-weather and operational-risk assistant

Start with METAR, TAF, SIGMET, PIREP and active advisory data, then translate it into the operation at hand. A dispatcher may care about ceiling and crosswind. A passenger cares whether a connection is becoming unrealistic.

TaskFuel already has raw aviation weather through 2s and StableTravel. SuVerse is another surveyed source; x402 Factory derives airport delay risk. The agent’s real job is to state the risk, the evidence behind it and when to check again.

6. A carbon-aware itinerary planner

Put fare, duration, stops and estimated CO₂e in the same comparison. MyCO2 Suite can estimate emissions from route and cabin; BRIJ returns emissions alongside its live airline offers. Your agent can apply a rule such as “prefer rail for short legs,” “avoid an extra stop unless it saves more than €200” or “show the lowest-emission option within two hours of the fastest trip.”

7. An airport-arrival brief

An arrival brief is a good first build because it replaces the six tabs travellers otherwise open after landing.

Start with the live flight and arrival terminal. Add destination weather, air quality, timezone, local time and sunset. Then check ground transport from the airport, allowing for the real arrival time rather than the schedule.

StableTravel or 2s for the flight and airport, OpenWeather for conditions and air quality, Timezone for local-time conversion, 2s for sunrise and sunset, and Mapbox for road, walking and travel-time comparisons. Confirming that an airport train still runs needs TransitPulse or another timetable source.

The result should be short enough to read while taxiing. For example:

You land at 18:42 local at Terminal 1. Light rain, 12°C; air quality is good. Sunset is 19:27. The fastest road route to the city centre is about 35 minutes. Check the airport train timetable before relying on the last service.

How to start without building every integration

Pick one decision you want the agent to make. “Track every flight” is vague. “Tell me when my pickup time needs to move by more than 15 minutes” is buildable.

Then give the agent the smallest set of sources that can answer it:

one source for the core flight or fare;
one genuinely independent source if failure would matter;
one contextual source, such as weather or ground transport;
a clear spending limit and a rule for when to run again.

TaskFuel helps your agent discover, evaluate and choose the right tool for the job. It can inspect each operation, dry-run the exact request to get a quote, and make the call with the relevant context. This lets your agent complete the task using the most suitable tools available.

Connect your agent and the first $5 is on us.

Ask it: Check whether LH400 is on time, tell me what the weather will be at arrival, and work out the fastest route from the airport to Manhattan after I land.

That is where flight data becomes useful: not another dashboard, but an answer that changes what you do next.

Catalog note: StableTravel and 2s are already available through TaskFuel for the aviation core. OpenWeather, Timezone and Mapbox cover useful context around the trip. The other providers are currently being added.

Introducing TaskFuel

Moritz Kaminski — Wed, 26 Aug 2026 07:09:28 +0000

Your agent can plan anything and reach almost nothing. So we gave it a balance and about a hundred services it can buy on its own: an inbox, a phone number, a computer by the hour, a letter in the actual mail.

TaskFuel is live today. The first $5 is on us.

Here’s why we built it.

Your agent can plan anything and reach almost nothing

Ask your agent to phone someone and tell you what they said. To make a real piece of music, not a description of one. To have an email address of its own, so it can send something and read the reply. To tell you what a flight costs right now, rather than what it cost whenever it was trained. To enrich a lead with the data sitting behind a social network’s login, or put an app on a server without you opening another account first.

All of that exists. But each one comes with a signup, another subscription, and a set of keys to look after. That’s an afternoon of work.

So you don’t bother. Your agent gets whatever toolbox you had the patience to build and works with that. Anything outside it can’t be reached, and you’re back in the loop.

How it works

TaskFuel gives your agent a budget and a catalog of a hundred specialist tools and services. It can ask what a particular call is going to cost, decide whether it’s worth paying for, and then go ahead and make it. You keep one account with TaskFuel, not one with every service in it.

Discover, evaluate, pay. That’s all it takes for an agent to get something done.

→ discover "music generation"
  Suno — POST /suno/generate-music

→ evaluate  {"customMode": false, "instrumental": true, …}
  $0.10  ·  nothing charged  ·  balance $5.00

→ pay
  track.mp3  ·  charged $0.10  ·  balance $4.90

What it can suddenly do

An inbox of its own runs $2 a month. A track from Suno, $0.10. A Linux VM at $0.01 an hour, with a public URL and a subdomain for $0.20 more.

Your agent can register a domain. It can send a physical letter.

Specialized web search across a dozen engines, company and people enrichment, SEC filings, market data, crypto prices, weather, maps, property records, translation, OCR, transcription, code execution, image and video generation, frontier models from half a dozen labs.

Any one of those is a convenience. What makes it different is chaining them into a whole workflow. An agent can enrich a lead from a network with no public API, spend half a cent confirming the business is real, write from its own address, and be there when the reply lands. Or register a domain, rent a machine for $0.01 an hour, and put the thing it just built on a public URL, all inside one run, without stopping once to ask you for a key.

Your agent doesn’t need a $99 seat to look up one company. It needs the lookup, whenever it comes up, for half a cent.

The obvious question

You’re about to hand a language model a spending limit, so it’s fair to ask how badly this can go.

Your agent spends against a balance and never touches your payment method. No single call can go above ten dollars, and you’ll be able to adjust that. Every cent is itemized as it happens. The few things that renew, an inbox among them, come out of the same balance as everything else, so if the balance is empty at renewal the inbox just lapses. An inbox nobody has touched for a month deletes itself anyway.

Worst case, something stops working. That’s the whole downside.

Why us

We’re the team behind Alby. We’ve spent years on payments, working towards a web where sending money is as easy as sending a message. This is that same work, pointed at the place it finally fits.

An agent’s problem usually isn’t intelligence. It’s reach. So we started with a catalog we picked by hand, then realised that finding a service doesn’t move anyone’s task closer to done. TaskFuel handles the buying too, and the guardrails that let you hand your agent a budget without sitting there watching it.

For the curious: how the money moves
Skip this bit if you don’t care.

There are a handful of open standards that let a service quote a price to a machine and get paid on the spot, no account and no checkout page. They’re a good thing. The catch is that there are several of them: x402, L402, MPP, with more arriving. Wire your agent up to a service yourself and you’re quietly betting on which one wins.

We don’t think that bet should be yours. We’re not adding another standard. TaskFuel sits in the middle, speaks all of them, pays whichever one a service asks for, and reports back to your agent in plain dollars.

What’s still rough

A hundred services is a thin slice of what’s actually for sale on the internet. If the thing your agent needs isn’t there, tell us. Adding services is the fastest thing we do, and right now one message is genuinely enough.

Discovery is keyword search. It’ll find you something that does the job, but it won’t tell you which of two options does it better, because we have no success rates, no ranking and no reputation layer yet. We know how we want to fix that. It’s next.

The part we’re actually excited about
Think about what your agent can do right now. That list got fixed on the day you set it up, and it doesn’t move until you go back and add something yourself.

We want it to grow without you. Every service we add is one your agent can already reach: no setup, no new account, nothing on your end at all. What it couldn’t do last month, it can do this month, while you weren’t looking.

Every service we add makes every agent out there a little more capable. That’s a good deal for everyone, and it’s why we’re doing this.

Come and try it

Connect your agent and the first $5 is on us. Go spend it.

And if you run a service you’d like agents to be able to buy, come talk to us.