The VEXYL Voice Gateway bridges SIP trunks, PSTN, Asterisk, FreePBX, business PBXs and WebRTC to the speech recognition, AI and voices you choose. Self-host it next to your PBX, or use it on VEXYL Cloud.
; Asterisk dialplan: hand the call to AI [voice-assistant] exten => _X.,1,Answer() same => n,Set(SESSION_UUID=${TRIM(${SHELL(uuidgen)})}) same => n,AudioSocket(${SESSION_UUID},127.0.0.1:8080) same => n,Hangup() # or place an outbound AI call POST /outbound/initiate { "phoneNumber": "+919876543210" }
AudioSocket over TCP straight from your dialplan, or a SIP trunk. Barge-in, DTMF and call transfers.
Accept calls from carrier trunks and PBXs, with or without registration. G.711 and Opus.
Panasonic KX-NS / KX-NSX, Avaya IP Office, Grandstream UCM and Bitvoice, as a trunk or an extension.
WebRTC voice on your website over a WebSocket, with the same agent as your phone line.
Start calls by API with per-call context and settings; follow them with lifecycle webhooks.
Pipe audio to a realtime speech-to-speech model for the most natural turn-taking.
Voice-activity detection, interruption handling, noise floor and silence settings you can tune.
Signed lifecycle events and per-call usage so you can bill and report accurately.
IP allow-lists, health probes, logs and production runbooks in the docs.
A modular pipeline that decouples telephony from intelligence. Swap any component without rewriting code.
Takes raw audio from Asterisk (AudioSocket over TCP), SIP trunks and PBXs, or web browsers (WebSocket).
Streams audio to speech recognition, sends text to the AI brain (LiteBot or your HTTP endpoint), and manages conversation state.
Synthesises the reply, caches repeated phrases, and streams audio back to the caller with smart buffering.
Bring your own keys and pay providers directly. Switch per deployment or per call.
Reminders, collections and surveys triggered by API, with context injected into the agent and no awkward silence when the customer says hello.
Standalone binary, set up for you.
curl -fsSL https://vexyl.ai/downloads/install_gateway.sh | sudo bash
Runs anywhere Docker runs.
docker pull vexyl/vexyl-voice-gateway:latest
High availability with auto-scaling.
kubectl apply -f vexyl-deployment.yaml
Full steps in the binary and Docker quick starts. Part of VEXYL Studio.
Same platform, two ways to run it. Start self-hosted for free, or let us run it for you on VEXYL Cloud.
Run the gateway and LiteBot on your own servers or VPC. Your audio, transcripts and keys never leave your infrastructure.
We run the gateway for you. Point a SIP trunk from your PBX at us, drop the browser widget on your site, and receive signed webhooks.
The VEXYL Voice Gateway connects phone calls and browser audio to AI. It takes audio from a SIP trunk, Asterisk, an IP PBX or a web page, sends it to a speech recognition provider, passes the text to the AI brain, and speaks the reply back with a text-to-speech voice, in real time.
Through AudioSocket over TCP, straight from your dialplan, or through a SIP trunk. Barge-in, DTMF and call transfers are supported.
Any SIP trunk that speaks G.711, Asterisk and FreePBX, and IP PBXs including Panasonic KX-NS / KX-NSX, Avaya IP Office, Grandstream UCM and Bitvoice, either as a trunk or as a PBX extension.
Yes. Start a call with POST /outbound/initiate and pass JSON context that is injected into the agent prompt. Webhooks report initiated, answered, completed and failed events. For CSV campaigns, use the LiteBot auto-dialer.
Either LiteBot, which gives you managed bots with prompts, tools, transfers and end-of-call hooks, or your own HTTP endpoint, such as an n8n or Flowise workflow or a custom backend.
Use the install script, Docker, or the standalone binary for bare-metal and air-gapped servers. Or skip installation and use the gateway on VEXYL Cloud.