A voice for
your tools.
An AI receptionist that welcomes callers, retrieves relevant information, transfers calls and takes messages.
Your phone line. Your models. Your knowledge.
PrismVox connects to your SIP trunk or PBX through Asterisk. An OpenAI-compatible model leads the conversation and uses the telephony tools, knowledge base and MCP servers you configure.
A conversation that can take action.
- Answer questions using approved documents.
- Transfer a call directly or introduce the caller to the recipient.
- Take a message and review conversation history.
- Make an outbound call with a specific mission.
Speech that fits your setup.
ElevenLabs is the default speech provider, with streaming recognition and synthesis. You can also connect your own STT and TTS services if they expose the documented compatible APIs. PrismVox does not install self-hosted speech engines.
External providers process the data you send them and may charge for usage. The code is open source; the telephony and AI services you choose are separate.
Independent, or connected to Prism.
In standalone mode, PrismVox has its own console, directory, knowledge and MCP tools. Connected to Prism, it can recognize callers from authorized profiles and give them access to the voice agent with their permitted knowledge.
Start on your own setup.
You need Docker, a SIP line or PBX, a tool-capable conversation model and a speech provider. The guides cover installation, every setting, external speech alternatives and remote administration.