Automate Azure Communication Calls with Java

Azure Communication Call Automation for Java is a server-side SDK for building call automation workflows: IVR systems, call routing, recording, and AI-powered interactions, authenticated via CallAutomationClientBuilder with DefaultAzureCredential or a connection string. It has five key classes: CallAutomationClient (make/answer/reject/redirect calls), CallConnection (actions on an established call), CallMedia (play audio, recognize DTMF/speech), CallRecording (start/pause/resume/stop), and CallAutomationEventParser (parse webhook events from ACS).

An outbound call targets a PSTN phone number with a source identity, caller ID, and a callback URL for events; an incoming call is answered using the incoming-call context from an Event Grid webhook plus a callback URL. Playing media supports text-to-speech (a TextSource with text and a named voice like en-US-JennyNeural) or a pre-recorded audio file from blob storage, targeted at specific participants. Recognizing input covers DTMF tones (a max-tone count, inter-tone timeout, stop tones, initial-silence timeout, and a spoken prompt) and speech (an end-silence timeout, a language, and a prompt), both started via callMedia.startRecognizing(). Recording starts against a server-call locator with a channel (mixed), content type (audio+video), and format (MP4), and can be paused, resumed, stopped, and downloaded once a RecordingFileStatusUpdated event fires. Participants can be added with an invitation timeout, and calls can be blind-transferred to another phone number. Webhook events are parsed with CallAutomationEventParser.parseEvents() into typed events - CallConnected, RecognizeCompleted (carrying DTMF tones or a speech result), PlayCompleted, CallDisconnected - dispatched by instanceof. Hanging up can end the call for all participants or drop only the local leg.

Use it when building server-side telephony workflows - IVR menus, call routing, recording, or AI-driven voice interactions - against Azure Communication Services, rather than a client-side calling UI SDK. Errors from the REST layer surface as HttpResponseException, and the skill's own examples branch on status code (404 for a call already ended, 400 for an invalid request), which you should handle explicitly rather than letting a stale call reference fail silently.

Input is call targets (phone numbers or communication user identities), media to play or recognize, and recording configuration. Output is call-connection state, recognized DTMF/speech results, recording files, and webhook events describing call lifecycle progress. Install it with:

<dependency>
    <groupId>com.azure</groupId>
    <artifactId>azure-communication-callautomation</artifactId>
    <version>1.6.0</version>
</dependency>

It's the azure-communication-callautomation client against Azure Communication Services, using Azure's neural text-to-speech voices for prompts, Event Grid for incoming-call notifications, and blob storage for both playing audio files and downloading call recordings.

Java developers building server-side telephony applications - IVR systems, call routing, recording, and AI-assisted voice interactions - on Azure Communication Services who need programmatic control over live calls.