LiveKit
Open-source real-time communication platform
✽
LiveKit is an open-source platform for building real-time audio, video, and data applications. Built on WebRTC, the platform provides a media server, client and server SDKs, and an agent framework for multimodal AI applications. LiveKit powers ChatGPT's Voice Mode and reached a $1B valuation in January 2026.2 The company is in the Founders, Inc. portfolio.1
Russ d'Sa, an entrepreneur who had worked at Twitter, 23andMe, and most recently as head of product at Medium, started LiveKit during the COVID-19 pandemic in the summer of 2021. When Clubhouse launched its audio-only social app in alpha, d'Sa built a "Clubhouse for companies" side project and collected a 1,300-company waitlist. Pinterest approached wanting 500 seats and asked where the audio data was going, which led d'Sa to conclude that no open-source infrastructure existed for real-time audio and video.9
D'Sa reached out to David Zhao, his former co-founder. Together they launched LiveKit as an open-source project in July 2021. D'Sa said the repository reached GitHub's top-10 trending list across all programming languages.9
Companies including Spotify, Oracle, and Reddit began using the self-hosted version and asked for a managed cloud offering. In December 2021, LiveKit announced a $7M seed round led by Redpoint Ventures with angel investors including Justin Kan, Robin Chan, and Elad Gil. The team had grown from 3 to 15 people, with over 100 projects using LiveKit and roughly 2,000 GitHub stars.3
The LiveKit team built a voice interface demo for the ChatGPT website on their own initiative. Four months later, OpenAI found the demo and reached out. In September 2023, LiveKit and OpenAI together launched ChatGPT Voice Mode.7
The LiveKit client SDK in the ChatGPT app captures speech and streams it over LiveKit Cloud to OpenAI's voice agent, which relays the audio to GPT-4o. The response streams back through the same path. The system uses WebRTC rather than WebSocket for the client-server connection because WebRTC handles packet loss more gracefully.6
D'Sa called LiveKit the "nervous system" connecting AI to the real world: "OpenAI is building a brain, AGI, and the input/output to that brain won't be a keyboard and mouse. It'll be vision and voice."8
LiveKit's Series A came in June 2024 at $22.5M, led by Altimeter Capital with Redpoint Ventures. Angel investors included Jeff Dean (Google), Aravind Srinivas (Perplexity), Amjad Masad (Replit), Guillermo Rauch (Vercel), and Elad Gil. At the time, 20,000 developers were using the platform, and LiveKit was being used by four of the ten most valuable private companies and six of the Fortune 50.5
In April 2025, LiveKit raised a $45M Series B at a $345M valuation, again led by Altimeter Capital with Redpoint Ventures and Hanabi Capital (Mike Volpi's fund). The company had grown to 100,000 developers. LiveKit Cloud was collectively handling over 3 billion calls per year, and a quarter of 911 emergency dispatch centers in the US were using LiveKit.107
The Series C followed in January 2026: $100M at a $1B valuation, led by Index Ventures with Salesforce Ventures, Hanabi Capital, Altimeter, and Redpoint. The LiveKit Agents framework was being downloaded over 1 million times per month, and the company had partnered with telephony carriers worldwide to connect to the public switched telephone network.24
LiveKit's architecture uses a Selective Forwarding Unit (SFU) that routes media streams from publishers to subscribers without decoding or re-encoding, keeping latency under 150 milliseconds.9 The platform is available both as self-hosted open-source software (Apache 2.0 license) and as LiveKit Cloud, a managed service.8
The product line includes three main components. LiveKit Server is the open-source media server (18,200 GitHub stars as of April 2026). LiveKit SDKs cover web and native clients (JavaScript, Swift, Android, Flutter, React Native, Unity, Rust) and server-side languages (Go, Node.js, Python, Ruby). LiveKit Agents is a framework for building real-time multimodal AI agents in Python and Node.js (10,100 GitHub stars).4
The company also offers Agent Builder for prototyping, Agent Observability for monitoring, a serverless agents runtime, and LiveKit Inference, a low-latency model gateway for voice AI. In December 2025, LiveKit released an SDK for the ESP32 microcontroller in partnership with Espressif Systems.4
Tesla uses LiveKit for voice AI across sales, support, insurance, and roadside assistance. Salesforce's Agentforce voice agents run on LiveKit. In iOS 18, LiveKit is embedded into FaceTime for emergency calls, streaming audio, video, and GPS data to 911 dispatch agents.84
Skydio uses LiveKit for police drone teleoperation, and Polymath Robotics uses it for remote operation of heavy machinery. Other customers include Spotify, Meta, Microsoft, xAI, Oracle, Adobe, Character AI, Coursera, and HeyGen.4