VoxLayer is a SaaS (or VAaaS?) platform that hosts conversational AI agents designed to live directly on your phone system. Instead of complex telco integrations, VoxLayer connects to your PBX as a standard SIP extension, allowing AI to answer calls just like a human team member.

Traditional IVRs (Interactive Voice Response) rely on rigid "press 1 for sales" menus. VoxLayer agents use natural language processing to understand intent. Customers can speak naturally, and the AI responds in real-time without the need for touch-tone menus.

No. VoxLayer is designed to be "plug-and-play" with your existing infrastructure. If your PBX supports standard SIP extensions (like Cisco, Avaya, 3CX, or Asterisk), you can add a VoxLayer agent in minutes.

VoxLayer is ideal for any business with high call volumes, including healthcare (scheduling), logistics (tracking), retail (orders), and financial services (account inquiries).

Yes. Since the AI is registered as an extension, it can place outbound calls through your PBX to handle appointment reminders, follow-ups, or automated notifications using your company's caller ID.

VoxLayer acts as a SIP User Agent. You simply create a new extension on your PBX and provide VoxLayer with the registration credentials (SIP Server, Extension/Username, and Password). Our platform then registers directly to your system.

No. Because VoxLayer initiates the registration to your PBX, it typically works through standard NAT/Firewall configurations. We also provide specific IP ranges for allowlisting to ensure a secure, stable connection. Our agents embed the WireGuard protocol for truly securing the connection if your firewall supports it.

Absolutely. You can register as many extensions as you need. Each extension can be mapped to a different AI persona, department, or language.

Yes. We support SIP over TLS for encrypted signaling and SRTP for the audio media, ensuring your internal phone credentials and conversations remain private. If your firewall supports

VoxLayer monitors registration status in real-time. If a connection is lost due to network jitter, our system automatically attempts to re-register. You can also monitor extension status directly from the VoxLayer dashboard.

VoxLayer is model-agnostic. You can choose from leading LLMs like GPT-4o, Claude 3.5, or specialized low-latency models we offer internally. We also offer a selection of high-fidelity Speech-to-Text (STT) and Text-to-Speech (TTS) engines.

Our platform features advanced Voice Activity Detection (VAD). If a caller interrupts the AI mid-sentence, the agent instantly stops talking and listens to the new input, creating a fluid, human-like cadence.

We optimize the entire stack—from SIP signaling to LLM processing—to keep response latency under 1 second. This prevents the awkward pauses often associated with older voice AI.

Yes. Since the agent is a registered extension, it can perform a standard SIP transfer (Blind or Attended) to move the caller to a human agent, another department, or a voicemail box on your PBX.

Language support and auto detection depends heavily on model choices for all three ASR, LLM, TTS. Our Whisper-based models support 99 languages, with varying degrees of accuracy based on available training data (best include English, Spanish, Italian, German, Portuguese, French, and Japanese), but don't support language auto-detection.

This depends on your PBX configuration. While a standard physical phone handles one call, VoxLayer can be configured to handle multiple concurrent "call appearances" on a single extension, or you can register a range of extensions to distribute load. Each potential call appearance and/or extension is billed as one agent instance, so if you anticipate handling 4 inbound calls at a time to the agent extension, you need four agent instances.

Simplicity. Registering as an extension means you don't have to manage complex SIP Trunking logic, DID routing, or specialized hardware. If your PBX can support a remote worker's softphone, it can support VoxLayer.

We operate today in Amazon's US-East data center with very high speed access to primary Internet exchange points in the area. In the future we will launch agent hosting nodes in other Amazon data centers around the world for better latency profiles in world-wide markets.

Yes, we offer enterprise-grade Service Level Agreements (SLAs) starting at 99.9% uptime.

We are SOC2 Type II compliant. For healthcare clients, we provide Business Associate Agreements (BAAs) to ensure HIPAA-compliant handling of all voice data and PII.

We use industry-standard encryption to store your PBX credentials, ONLY for the agent instance. These are never accessible to unauthorized personnel.

By default, we provide real-time processing without storage. You can opt-in to call recording and transcription for training purposes, or choose a "zero-retention" mode for maximum privacy.

Our agents are wrapped in a proprietary security layer that filters inputs for malicious instructions, ensuring the agent follows your specific business scripts and doesn't deviate from its intended role.

Our base platform fee is $99/month (or $999/year if billed annually) per agent instance.

VoxLayer offers a hybrid "Bring Your Own Key" (BYOK) or "Turnkey" model:

Internal Models: If you use VoxLayer's optimized internal models for transcription (ASR), reasoning (LLM), or speech (TTS), they are billed as individual add-ons.

Public Models: You can use models from OpenAI, Anthropic, or Google (and others) for free on our platform, provided you supply your own API keys and pay those providers directly for your token usage.

Yes. If you choose to use our curated Internal Models, you can bundle your agent and its options into a fixed monthly or annual price. A fully-loaded agent typically costs under $250/month, giving you total budget predictability without worrying about fluctuating token costs.

Our Tool Marketplace features both free and premium integrations. While many core tools are included, some advanced third-party tools (like specialized CRM syncs or industry-specific data connectors) may carry their own monthly or annual subscription fees.

Our pricing is primarily platform-based. You pay the base fee for each agent instance, and your total monthly cost scales based on the specific models and marketplace tools you activate for your agents. If you anticipate an agent extension handling four simultaneous calls, you need four agent instances.

The app is designed to be your agent's "ears" while you're on the move. It runs in the background of your mobile device and stays ready to assist you the moment you need it.

Yes, but with more freedom. You can define your own custom wake word (e.g., "Hey Vox" or "Listen up, Agent"). The app listens for this trigger and immediately opens a private voice channel to your personal agent.

Absolutely. The app is optimized for use with in-ear headphones. While you speak to the agent, the responses are played directly into your ears, ensuring your digital assistant's feedback and your personal data stay private, even in public spaces.

Every personal agent on VoxLayer includes an OpenClaw instance. OpenClaw is the "brain" that actually executes tasks. While VoxLayer handles the voice interface and telephony, OpenClaw manages the logic, memory, and tool execution—like a high-powered, open-source version of a personal chief of staff.

All configuration is done via the VoxLayer Web Interface. There, you can select which LLM "brain" to use, pick a voice, and configure your OpenClaw integrations (like your calendar, email, or task managers). Once saved, those changes are instantly live on your mobile app.

Because of the OpenClaw backbone, you can do far more than just ask questions. You can tell your agent to:

"Schedule a meeting with the dev team for Thursday."

"Set an alert to notify me if my server latency exceeds 200ms."

"Draft an email summary of my last three calls and send it to my manager."

"Remind me to follow up on the Smith contract when I get back to the office."