Vocode
Open-source Python framework for building programmable voice agents over phone, web, and Zoom.
About
Vocode is an MIT-licensed open-source framework for building voice-based LLM applications in Python. Its design is deliberately modular: developers compose their own stack by plugging in a speech-to-text provider such as Deepgram, an LLM such as GPT-4, and a speech synthesis provider such as ElevenLabs, then wire them through Vocode's streaming and turn-based conversation abstractions with endpointing, interruption handling, and actions. It supports inbound and outbound phone calls, web voice experiences, and Zoom meetings, plus a hosted API and Node/React SDKs for teams that do not want to self-host.
The project was an early reference implementation in the voice-agent space and still shows meaningful adoption, with roughly 3,800 GitHub stars and 650 forks as of October 2026. Momentum has clearly slowed, however: the repository's last commit was in November 2024 and the maintainers have publicly said they are looking for community maintainers. Treat it as useful reference architecture and a workable self-hosted starting point rather than an actively developed production platform; newer open-source alternatives like Pipecat and LiveKit Agents have taken the maintenance lead.
What it can do
- Modular STT, LLM, and TTS pipeline with streaming and turn-based conversation modes
- Inbound and outbound phone calls with configurable prompts, voices, webhooks, and actions
- Web voice experiences and Zoom meeting participation
- Self-hosted Python/Docker deployment or hosted API
- MIT-licensed core usable commercially at no license cost
Setup
- Clone the vocodedev/vocode-python repo or pip-install the package, and bring API keys for your chosen STT, LLM, and TTS providers.
- Configure an agent: pick a conversation type (streaming or turn-based), set the prompt, voice, and interruption behavior.
- For telephony, attach a phone number via the hosted service or wire up your own Twilio account and Redis.
- Deploy the Python service (Docker images are provided) and connect it to your web app or Zoom flow via the SDK.
Prerequisites
Python environment; API keys for transcription, LLM, and TTS providers; Twilio account and Redis for self-hosted telephony.
Use cases
Works with
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: ~3,800 GitHub stars and 656 forks (GitHub API, Oct 2026); MIT-licensed open-source core with hosted API; listed as a standard reference architecture across voice-agent comparisons (aiagentstore, aibucket, Dasha). Strengths: genuinely open and self-hostable, modular provider design, no license cost. Docked for: (1) repository effectively unmaintained since November 2024 with maintainers seeking community handoff; (2) hosted pricing not publicly documented and telephony requires paid upgrade. License: MIT (core).
★ 3,804 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 3.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · ~3,804 GitHub stars, 656 forks (GitHub API, Oct 2026)
- · Last commit Nov 2024; maintainers publicly seeking community handoff
- · Modular open-source voice-agent framework; hosted API available
- · Listed as reference architecture across multiple voice-agent comparisons
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.