Zoom Buys Onyx to Add Open-Source Search to Zoom AI

Zoom has acquired Onyx, an open-source platform for enterprise search and knowledge management, and is folding it into Zoom AI Services. In its newsroom announcement, Zoom said the deal gives its AI agents the organisational context to “ground decisions and agent actions”, and that Onyx, founded in 2023, connects to more than 50 enterprise applications and is fully self-hostable. Zoom put no price on the acquisition.

Onyx confirmed the move in its own post, leading with reassurance to its users: “nothing about how you use Onyx is changing. Same product, same team, same service.” Co-founders Chris Weaver and Yuhong Sun are joining Zoom, which says Onyx “will remain a standalone platform and it will stay open source, model-agnostic, app-agnostic, and self-hostable”. The project remains public on GitHub, and Zoom cited a community of more than 30,000 developers and customers including NASA and Hudson River Trading processing over a million queries a week.

In brief

Zoom has bought Onyx, an open-source enterprise search platform that connects to 50-plus apps and can run entirely on a customer’s own infrastructure.
Zoom says Onyx stays standalone, open source, model-agnostic and self-hostable; co-founders Chris Weaver and Yuhong Sun join Zoom.
No price was disclosed. The buy feeds Zoom’s “Conversation to Completion” push to give AI agents organisational context.

Why Zoom wants a context layer, not another chatbot

The logic here is about grounding, not chat. A capable model can answer almost anything about the world yet knows nothing about a given company’s policies, tickets or customer history, and an AI agent that acts without that context is a liability. Onyx is a retrieval layer that pulls from the systems where that knowledge actually lives and serves it to both people and agents on demand – exactly the gap Zoom described when it first set out its unified context layer to let AI agents finish what meetings start. Buying Onyx rather than building it buys Zoom time and a working codebase with an existing enterprise footprint. It also slots into a broader land-grab for the enterprise knowledge surface, the same instinct behind rivals pushing assistants deep into day-to-day work, as when Uber rolled out ChatGPT Work to more than 30,000 employees. For a company whose core product is the meeting, owning the layer that turns what is said into what gets done is a plausible way to stay central as agents take on more of the follow-through.

The open-source promise is the part to watch

Acquiring an open-source project always raises the same question, and the answer rarely arrives on day one. Zoom and Onyx have both pledged that the platform stays open source, self-hostable and model-agnostic, which matters to the data-sovereignty and on-premises customers who adopted it precisely because they could run it themselves. But a pledge is only as good as its follow-through: the things to watch are the public release cadence on GitHub, whether self-hosted builds keep pace with whatever Zoom ships to its own cloud, and how the licence is handled over time. The record for AI acquisitions is mixed on exactly this point, with acquirers often absorbing a team into a closed roadmap, as happened when OpenAI bought Orchestra Bio and its agents. Onyx’s distinctiveness is its openness; dilute that and Zoom has paid, at an undisclosed price, for a team and a brand rather than a movement. The first real signal will be whether the next few Onyx releases land in the open repository as readily as they did last month.