canonical: https://jentic.com/apis/messagebird.com

# MessageBird APIs

MessageBird provides messaging and text-understanding APIs for building customer communication flows. The SMS API sends transactional messages and manages sent-message records over MessageBird's global mobile network, while a set of natural-language APIs classify intent, detect language, extract named entities, and predict grounded FAQ answers against curated datasets. Together they let a workflow send outbound SMS and read the meaning of the short text that flows back, without hand-rolling clients for each host. All five specs are Jentic-generated because MessageBird does not publish OpenAPI definitions for these products.

## For AI agents

An agent can send SMS messages and reconcile or delete the resulting records, then classify inbound message intent, detect the language of a phrase, extract structured entities, and match a question to a curated FAQ dataset. This lets one agent both dispatch outbound texts and interpret the free-text replies that come back.

## Scope

Use for: Sending and managing SMS messages, and classifying, detecting the language of, extracting entities from, or matching FAQ answers to short customer text

Not supported:
- voice calls
- WhatsApp messaging
- free-form text generation
- dataset training
- long-form documents
- message translation

## APIs

| API | Category | Endpoints | Description |
| --- | --- | --- | --- |
| FAQ | ai-ml | 4 | Predict the best FAQ answer to a user question against a curated MessageBird dataset across multiple languages. |
| Intent | ai-ml | 4 | Classify customer messages into trained intent labels by calling a MessageBird Intent dataset, with confidence scores returned per prediction. |
| MessageBird SMS API | communications | 4 | Send SMS messages and inspect or delete previously sent message records via MessageBird's REST surface, in 4 endpoints. |
| LanguageDetection | ai-ml | 2 | Detect the language of a phrase and return a language code so an agent can route, translate, or classify multilingual messages. |
| NamedEntityRecognition | ai-ml | 2 | Extract structured entities (names, places, dates, organisations) from a customer message so an agent can fill slots without keyword parsing. |

## Which API to use

| Need | API | Why |
| --- | --- | --- |
| Send a transactional or OTP text to a phone number | messagebird-sms-api | The SMS API is the only one here that delivers messages; the NLP APIs only analyze text. |
| Decide what an inbound message is asking for | messagebird-intent-api | Intent returns a trained label with a confidence score; NER returns entities and FAQ returns a canned answer, which answer different questions. |
| Extract specific values like dates, names, or places from a message | messagebird-ner-api | NER returns typed entity slots; Intent only returns the overall label and does not surface individual values. |
| Answer a support question from a vetted knowledge set | messagebird-faq-api | FAQ scores a question against a curated dataset; Intent routes by category rather than returning an answer. |
| Identify which language a phrase is written in before routing | messagebird-language-detection-api | Language Detection returns a language code with no training step; Intent and NER assume the locale is already chosen. |

## Cross-API use cases

### Detect, classify, and answer an inbound message

When a customer reply arrives, detect its language, classify what it is asking for, and match it to a vetted FAQ answer. Language Detection picks the locale, Intent returns the labelled category with a confidence score, and FAQ returns a grounded answer the agent can send back or escalate on low confidence.

Example prompt: Call predict-language to get the locale, POST the phrase to the Intent predict operation, then call the FAQ predict-answer operation on the matching dataset and reply if confidence clears the threshold.

### Slot-fill a booking from a text and confirm by SMS

Extract the date and location from an inbound booking request, act on the structured values, then send a confirmation SMS. NER turns the free text into typed entities and the SMS API dispatches the confirmation with the returned message ID stored for auditing.

Example prompt: POST the message to predict-entities, use the returned date and location to complete the booking, then POST a confirmation to /messages and store the message ID.

### Route an inbound SMS reply to the right workflow

For replies to outbound SMS campaigns, classify the intent of each reply and forward it to the correct queue or automated action. The SMS API surfaces the sent-message records for reconciliation and Intent labels the inbound reply so the next step is chosen without keyword matching.

Example prompt: List sent messages with GET /messages to reconcile the thread, then POST the inbound reply to the Intent predict operation and branch on the returned label.

## Why Jentic

- **Setup:** Wiring these MessageBird APIs by hand means attaching your AccessKey to requests against several different hosts (rest, intent, faq, language-detection, and ner) and handling routing and retries per host. With Jentic you install once, add the APIs you need from the Jentic directory, store the AccessKey once, and your agent calls them.
- **Permission scoping:** You choose which operations your agent may call across the five APIs, so your own rules can grant SMS sending and inbound-message analysis while leaving message deletion or dataset reads out of reach. Where a dataset id appears in the path, a rule can pin the agent to a single dataset.
- **Credential handling:** Your MessageBird AccessKey is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context, and the same key covers all five APIs.
- **Discovery method:** Agents search the Jentic directory by intent such as 'send an SMS' or 'classify a customer message', and Jentic returns the matching operation with its input schema so the agent calls the right endpoint across the MessageBird surfaces without browsing reference docs.

## Related vendors

- **Twilio** — Competing global SMS gateway with broader voice and conversation channels alongside messaging.
- **Sinch** — Global SMS and CPaaS provider with a similar transactional messaging surface.
- **Plivo** — Cloud SMS and voice API with worldwide reach and a comparable REST surface.
- **Cohere** — Hosted NLP API for building intent classifiers and entity extraction outside a messaging stack.

## FAQ

### Which MessageBird APIs are covered here?

Five: the SMS API for sending and managing text messages, plus four short-text NLP APIs for intent classification, language detection, named entity recognition, and FAQ answer prediction. The SMS API handles delivery and the other four interpret inbound text.

### Do these APIs share one set of credentials?

Yes. Every MessageBird API here authenticates with the same AccessKey in the Authorization header, so a single credential works across all five surfaces. You store that AccessKey once in your own Jentic One instance.

### How do the messaging and text-understanding APIs work together?

The SMS API sends outbound messages, and the Intent, NER, Language Detection, and FAQ APIs read the meaning of the free text that comes back. A typical flow sends a notification, then detects language, classifies intent, extracts entities, and answers or routes the reply.

### Why are all the specs Jentic-generated?

MessageBird does not publish OpenAPI specifications for these products. Jentic generates and maintains the specs so agents can call each API through structured tooling, and validates them against the live endpoints.

### When should I use the FAQ API versus the Intent API?

Use FAQ when you want a grounded answer scored against a curated question-and-answer dataset. Use Intent when you only need a category label to route a conversation. They are often chained: classify intent first, then predict the FAQ answer for that topic.

### Can these handle multilingual traffic?

The NLP APIs each expose a supported-languages lookup, and the Language Detection API returns a language code you can use to gate or route the other calls. Query the supported-languages endpoint at runtime rather than hardcoding the set, since coverage changes.

### Are these built for long documents?

No. The NLP APIs are tuned for the short utterances seen in SMS, WhatsApp, and chat traffic. For long-form document analysis, evaluate accuracy on representative samples or pair them with a general-purpose NLP model.