Vocode
Python library for building voice-based LLM agents that run on phone calls, Zoom and system audio.
Vocode streams speech-to-text, an LLM and text-to-speech into inbound and outbound phone calls.
Vocode (vocode-core) is an open-source Python library for real-time streaming conversations with LLMs that can be deployed to phone calls and Zoom meetings. It suits developers prototyping voice agents, though the project asks for community maintainers.
What it is
You assemble a streaming conversation from a transcriber, an LLM agent and a synthesizer, then connect it to your system audio, an inbound or outbound phone number, or a Zoom dial-in. It has built-in integrations with services such as Deepgram, AssemblyAI, OpenAI, Anthropic, ElevenLabs and Azure.
Who it's for
It suits developers who want a ready telephony-oriented voice agent library and are happy to maintain their own deployment. It does not suit anyone who needs a maintained, supported product, because the README says the team is looking for community maintainers.
Key features
Streaming conversations
Real-time speech-to-LLM-to-speech loop with interruption handling and endpointing.
Phone call agents
Set up a number that answers with an LLM agent, or send outbound calls managed by one.
Zoom dial-in
Includes code to dial into a Zoom call.
Many provider integrations
Transcribers, LLMs (OpenAI, Anthropic) and synthesizers such as ElevenLabs, Azure, Google, Cartesia and Rime.
Actions
External actions, call transfer, wait actions for IVR navigation and phrase triggers.
Langchain agent
An outbound call can be used inside a Langchain agent.
Up and running in 3 steps
Our honest verdict
What's great
- Telephony is a first-class use case: inbound and outbound calls are documented in the open-source docs.
- Wide list of speech and voice providers, including local options like Whisper.cpp, Coqui and Bark.
- pip install vocode is enough to run the quickstart with microphone and speaker.
- MIT licence; the 0.1.113 notes say functionality is no longer gated behind the hosted API.
Where it falls short
- The README says the team is actively looking for community maintainers, and the latest releases on the page are alpha pre-releases of 0.1.114.
- Telephony needs your own provider accounts plus transcriber, LLM and voice API keys, and the code to host the server.
- The repo's roadmap and docs are separate pages, so check them for supported features before committing.
- It is a Python library with no visual builder.
Pricing & license
Vocode is free under the MIT license; the only cost is running it yourself plus the transcription, LLM and voice services you connect. The README links a dashboard at app.vocode.dev but no prices were read.
Open remote AI-training roles
768 open remote AI-training roles, updated November 14, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is Vocode still maintained?
The README says it is actively looking for community maintainers, and the latest listed releases are 0.1.114 alphas after 0.1.113.
Can it make phone calls?
Yes, inbound and outbound calls are listed features, via telephony providers documented in the docs.
Is the hosted API required?
No. The 0.1.113 release notes say functionality is no longer gated behind the hosted API.
