By Hamza Faisal
Deepgram is acquiring Daily, the company behind Pipecat, one of the most widely used open-source frameworks for building voice AI agents. The voice AI infrastructure company announced the deal on October 29 at its inaugural Deepgram Speak event in San Francisco, pitching the combination as the first complete, open, production-grade platform for conversational voice AI.
Pipecat is the piece of plumbing that lets a voice agent actually hold a conversation. It manages voice and video streams and stitches together the pipeline behind them — speech-to-text, a large language model, and text-to-speech — while keeping latency low enough that the exchange feels live rather than turn-based. Daily, its maker, also provides the hosted infrastructure for voice and video calls of any kind.
Open source stays open — with a foundation to prove it
The question hanging over any acquisition of an open-source project is whether the new owner will quietly close it off. Deepgram's answer is unusually explicit: Pipecat will remain fully open source and free to all developers under its existing BSD V2 licence, and it will stay vendor-neutral, letting developers mix models, tools, network transport and APIs from different providers.
To back that promise, Deepgram and many of the organisations that contribute to and use Pipecat are establishing an independent Pipecat Foundation, and Deepgram says it will invest in the framework's continued development. Daily's team is joining Deepgram, with co-founder Kwindla Hultman Kramer becoming Head of Open Source. Kramer said keeping Pipecat open was fundamental to the decision to join Deepgram, favouring chief executive Scott Stephenson's commitment to keeping it open source and vendor-neutral over other acquisition interest.
One stack, from speech to conversation
For Deepgram, the deal fills the gap between its speech models and a finished voice agent. Daily's global real-time network, which connects end users to voice agents with ultra-low latency and high reliability, will underpin a new Voice Agent Platform alongside Deepgram's own models. The pitch is that developers can build and scale reliable agents while keeping control over their architecture and model choice, rather than being locked into a single vendor's stack.
Alongside the acquisition, Deepgram is launching the Real-Time Context Standard, an open protocol developed with the Pipecat team that carries context across an entire conversation — metadata about the audio itself, instructions telling a language model how to use that context, and guidance on how a spoken response should sound. The idea is that an agent should react not just to the words spoken, but to how they were said.
Why now
The voice AI sector is consolidating fast as the technology moves from demos to production systems handling real customer calls. Deepgram reached unicorn status earlier this year with a $130 million Series C round, while Daily has raised more than $62 million from investors including Tiger Global, according to PitchBook data. The companies did not disclose a purchase price. Incidentally, both were in the same Y Combinator batch in 2016 — a shared history Stephenson pointed to as evidence that the two engineering cultures match.
The deal will ultimately be judged on the foundation's independence and on whether Pipecat's neutrality survives inside a company selling its own speech products. Stephenson's stated ambition is to open the ecosystem so that everybody can build voice agents. If developers keep trusting Pipecat as neutral ground, Deepgram gains the orchestration layer of voice AI without having to win it by lock-in. If that trust erodes, the open-source community that gives the framework its reach could drift elsewhere — and with it, much of the value Deepgram just bought.
