Deepgram Acquires Daily, Maker of the Open-Source Voice Agent Framework Pipecat: Pipecat Stays Open Source and Vendor-Neutral, Under an Independent Foundation Whose Steering Committee Includes Competitors

The deal

Voice model company Deepgram announced on October 29 that it is acquiring Daily. Daily built realtime audio and video infrastructure, running specialized compute clusters in many locations and giving developers libraries to embed audio and video in web, desktop and mobile apps. From 2023 it started using that infrastructure for realtime voice agents, which is where the open-source Pipecat framework came from. Terms were not disclosed.

In a company blog post, Daily co-founder Kwindla Hultman Kramer laid out the arrangements. The Daily team joins Deepgram and keeps working on Pipecat and the global network infrastructure. For existing customers "nothing will change": Deepgram will sell, support and develop all of Daily's products, including WebRTC, telephony and cross-platform client libraries. Deepgram is adopting Pipecat as the core of its new voice AI products and services. And Pipecat stays open source and vendor-neutral, with an independent Pipecat Foundation managed by a steering committee that represents the whole realtime AI ecosystem, including companies that compete with Deepgram.

What Pipecat is

Pipecat ties together the stages of a voice agent pipeline: streaming data between speech-to-text, the language model and text-to-speech, plus model switching, low-latency network transport, turn detection, interruption handling, context management and async tool calls. It was fully open-sourced in 2024, including the core orchestration framework, client SDKs for every platform, the RTVI messaging standard, the Smart Turn turn-detection model and the Pipecat Flows context engineering library. According to Kramer, Pipecat now has more than 200 integrations and almost 400 contributors, its core was built by 18 engineers, and customer support voice agents were its first breakout use case. The two companies go back a long way: they were in the same 2016 Y Combinator batch, and Daily's audio-video APIs have used Deepgram for transcription since 2019.

TechCrunch, citing PitchBook, reports that Daily has raised more than $62 million; Deepgram raised a $130 million Series C earlier this year, reaching unicorn status. Deepgram also released the Real-Time Context Standard (RCS), an open protocol co-developed with the Pipecat team for passing metadata from the audio, such as tone, through a voice agent, telling the language model how to use those signals, and guiding how the response should sound.

What it means for developers

In voice agents, model makers (Deepgram, ElevenLabs and others) and orchestration and transport infrastructure (LiveKit, Vapi, Pipecat and others) used to be separate layers; this deal puts one of the most widely used open-source orchestration frameworks in the hands of a model maker. Deepgram CEO Scott Stephenson told TechCrunch he wants developers to use the best model for each use case rather than being tied to Deepgram. Teams using Pipecat won't feel a change in the short term; what's worth watching is who sits on the foundation's steering committee and whether Pipecat's integrations with other model vendors keep getting equal care.