canonical: https://jentic.com/apis/amadeus.com/amadeus-flight-delay-prediction

# Amadeus Flight Delay Prediction

The Amadeus Flight Delay Prediction API estimates the probability that a given scheduled flight will be delayed and assigns it to one of four delay segments - under 30 minutes, 30 to 60, 60 to 120, or over 120 minutes. The single GET endpoint takes the carrier code, flight number, scheduled departure date and time, origin, destination, duration, and aircraft code, and returns probability scores per segment. Travel apps use it to set traveller expectations, suggest tighter or looser connection times, and prioritise rebooking outreach.

## For AI agents

Estimate the probability a scheduled flight will be delayed by 30, 60, or 120+ minutes so an agent can warn travellers, adjust connection planning, or trigger proactive rebooking.

## Scope

Does not return live flight status, actual delay minutes, or rebooking options - use only to estimate delay-segment probabilities for a scheduled flight.

## Capabilities

- Estimate probability of delay across four segments (under 30 min, 30-60, 60-120, 120+) for a given scheduled flight
- Score upcoming flights in a traveller's itinerary by likely on-time performance
- Flag connecting itineraries with high delay probability on the first leg
- Trigger proactive traveller communications when a high-delay-probability flight is approaching
- Compare expected reliability across alternative carriers on the same route

## Use cases

### Proactive Delay Notifications

Travel apps run a daily job over upcoming bookings, call /travel/predictions/flight-delay for each segment, and email or push-notify travellers whose flight has a high probability of falling into the 60-120 or 120+ delay band. This gives passengers time to leave for the airport later or arrange alternative ground transport without the integrator needing real-time operational data feeds.

Example prompt: For each booking departing in the next 24 hours, call /travel/predictions/flight-delay and send a notification when the combined probability for the 60-120 or 120+ band exceeds 25 percent.

### Tighter Connection Risk Scoring

Self-service booking flows score the first leg of any connecting itinerary with the delay endpoint and surface a warning when the layover is short relative to the predicted delay range. The scoring takes a few hundred milliseconds per leg and can be cached for an itinerary's lifetime, so the impact on checkout latency is minimal.

Example prompt: Given a two-leg itinerary with a 50-minute layover, score the first leg's delay probability and return a warning if the 30-60 or 60-120 band combined exceeds 30 percent.

### Carrier Comparison on a Route

Corporate travel managers compare expected reliability of competing carriers on the same city pair by calling the prediction endpoint for representative future flights from each carrier. Aggregated over a week of departures, the resulting probabilities feed a simple reliability score used in supplier negotiations.

Example prompt: Call the delay-prediction endpoint for one representative flight per carrier on JFK-LHR for the next 7 days and rank carriers by expected probability of an under-30-minute delay.

### Agent-Driven Itinerary Risk Check via Jentic

An AI travel concierge calls Flight Delay Prediction through Jentic for each leg of a customer's upcoming trip and includes the delay risk in the daily briefing. Jentic resolves OAuth and parameter shaping, so the agent only describes the intent and supplies flight metadata.

Example prompt: Call jentic.search('predict Amadeus flight delay'), load the operation, and execute it with carrierCode, flightNumber, scheduledDepartureDate, scheduledDepartureTime, departureAirportCode, arrivalAirportCode, aircraftCode, and duration to return per-segment probabilities.

## Key endpoints

| Method | Path | Description |
| --- | --- | --- |
| GET | /travel/predictions/flight-delay | Predict delay segment probabilities for a scheduled flight |

## Key resources

- **Flight Delay Prediction** — Single GET endpoint that takes flight metadata and returns per-segment delay probabilities.

## Why Jentic

- **Setup:** Wiring the Flight Delay Prediction API by hand means handling the Amadeus OAuth 2.0 client_credentials exchange, refreshing the bearer token, and assembling the query parameters for GET /travel/predictions/flight-delay against the correct Amadeus host and version. Through Jentic you install once, import Flight Delay Prediction from the API Directory, store the Amadeus client credentials once, and your agent calls it.
- **Permission scoping:** The route, date, and carrier details travel as query parameters, not as a resource id in the URL path, so scoping is by operation. You limit the agent to the operations it needs, such as reading the delay-segment probabilities for a scheduled flight, and it cannot reach other Amadeus operations unless you add them.
- **Credential handling:** Your Amadeus client_id and client_secret are stored once, encrypted, by your own Jentic One instance, which runs the client_credentials exchange and injects a fresh bearer token at execution time. The raw secret never enters the agent's prompt, logs, or context.
- **Discovery method:** Agents search Jentic by intent such as 'predict how likely this scheduled flight is to be delayed', and Jentic returns the /travel/predictions/flight-delay operation with its input schema so the agent supplies the required query parameters without reading the Amadeus docs first.

## Related APIs

- **On-Demand Flight Status** — Returns the live operational status of a scheduled flight, complementing the predictive score.
- **Airport On-Time Performance** — Predicts on-time performance at an airport level rather than a specific flight.
- **Flight Choice Prediction** — Scores offers by likelihood the traveller will pick them; pair with delay risk to surface reliable, attractive options.
- **Trip Purpose Prediction** — Different prediction model focused on whether a trip is business or leisure rather than delay risk.

## FAQ

### What authentication does the Flight Delay Prediction API use?

Amadeus Self-Service OAuth 2.0 client_credentials. Exchange client_id and client_secret at https://api.amadeus.com/v1/security/oauth2/token for a bearer token and pass it in the Authorization header. Through Jentic, the credentials are kept in the encrypted vault and the bearer token is rotated automatically before each prediction call.

### Which fields do I have to send to get a prediction?

All eight query parameters are required: originLocationCode, destinationLocationCode, departureDate, departureTime, arrivalDate, arrivalTime, aircraftCode, carrierCode, flightNumber, and duration. Missing any of them returns a 400 from /travel/predictions/flight-delay.

### What are the rate limits for the Flight Delay Prediction API?

Amadeus Self-Service Test environment caps prediction endpoints at 10 transactions per second per API key with a fixed monthly quota; Production limits scale with your Self-Service plan tier. Bearer tokens expire after roughly 30 minutes and must be refreshed.

### How do I score a flight's delay risk through Jentic?

Run pip install jentic, then call jentic.search('predict Amadeus flight delay'), load the operation, and execute it with the eight required parameters. Jentic handles the OAuth exchange and calls /travel/predictions/flight-delay, returning the delay-segment probabilities.

### Does the API tell me the actual delay in minutes?

No. It returns a probability per segment (under 30 min, 30-60, 60-120, 120+). To translate into an expected value, multiply each midpoint by its probability and sum, but treat the output as a risk band rather than a precise prediction.

### Is the Flight Delay Prediction API free?

The Test environment is free with limited monthly quota. Production access is metered per transaction under an Amadeus Self-Service plan; pricing is published on the Amadeus for Developers portal.

### Can I limit what my agent is allowed to do with the Flight Delay Prediction API?

Yes. Because you self-host Jentic One, your own rules decide which operations and credentials the agent can use, and scoping here works at the operation level since the route, date, and carrier details travel as query parameters rather than a resource id in the URL path. You can allow the agent only the read operation that returns delay-segment probabilities for a scheduled flight from GET /travel/predictions/flight-delay, and it cannot reach any other Amadeus operation unless you add it. The Amadeus client_id and client_secret stay stored and encrypted in your own instance, which runs the OAuth exchange and injects a fresh bearer token at call time, so the raw secret never enters the agent's prompt or logs.