Onyx’s Open-Source AI Is Fixing Enterprise Search That’s Broken

Most companies shouldn’t build their own foundation models, at least that’s what Poolside CEO Jason Warner believes. But Onyx’s co-founder and co-CEO, Chris Weaver, sees things differently. He envisions Onyx as a foundational tool for enterprises looking to build internal search products, tackling a challenge many companies still struggle with: finding the right information when they need it.

The enterprise search space is already fiercely competitive, with players like Glean and Swirl leading the charge. Yet, despite the rapid advancements in AI, businesses continue to wrestle with the complexity of extracting value from their internal data. Employees waste time searching across vast company repositories like Salesforce, GitHub, Google Drive only to come up empty-handed. Onyx aims to fix that.

Weaver and his co-founder, Yuhong Sun, weren’t strangers to startup failures. Their earlier ventures, a live stats tracker for Twitch streamers and a keyboard comparison site never took off. But the duo saw a recurring issue in their engineering roles: information was scattered, hard to find, and even harder for new employees to navigate.

“We knew where things were roughly, but it was still kind of hard, [and] new people just couldn’t find anything,” Weaver told TechCrunch.

In 2023, they launched an open-source project, Danswer (a blend of “deep” and “answer”), and quickly gained traction. Companies wanted to use it and, unexpectedly, they wanted to pay for it.

“Ramp was actually one of the early teams that found us,” Sun said. “At the time, we didn’t have any way for them to pay us or anything… But people really wanted to pay for our project.”

That realization sparked Onyx’s transformation from an open-source tool into a full-fledged business.

Unlike many enterprise SaaS companies, Onyx leans into an open-source model, betting that transparency and accessibility will accelerate adoption. It’s already working with major enterprises like Netflix, Ramp, and Thales Group.

The startup promises a near-instant setup, integrating with over 40 internal data sources within just 30 minutes. Its enterprise users can opt for additional security and encryption features as part of a paid tier. Unlike traditional search tools that merely retrieve information, Onyx employs AI to refine and validate results, ensuring that responses are accurate, complete, and enterprise-ready.

Onyx, has adopted 37,000 users at the University of California, San Diego and is making this possible by enabling human-in-the-loop agent workflows that automate routine tasks.

One application of this technology is its integration with Ramp, allowing AI to efficiently process workflows that require both automation and human oversight. By leveraging Onyx and Ramp together, AI systems can streamline tasks, reducing manual effort while maintaining accuracy and adaptability.

“Open source is really the only way for this type of solution to scale out and get the momentum into every single business in the world,” Weaver said.

But it won’t be an easy fight. The company faces stiff competition, not just from startups like Glean but also from in-house AI tools built by companies like Klarna, whose internal search and chatbot tool, Kiki, serves a similar purpose.

Weaver remains confident, arguing that building a search tool from scratch is an enormous challenge. Onyx provides the foundational layer, eliminating that complexity.

And the demand is clear.

“We’ve seen the usage grow explosively,” Sun said. “We hit a peak of over 160,000 messages in a single week. We are really hoping to lean into that organic growth and hopefully all the teams in the world will use Onyx one day.”

The startup recently closed a $10 million seed round co-led by Khosla Ventures and First Round Capital, with backing from notable names like:

• Gokul Rajaram (ex-Coinbase, Pinterest board member)
• Arash Ferdowsi (Dropbox co-founder)
• Amit Agarwal (ex-Datadog Chief Product Officer)

With strong early adoption, open-source momentum, and a growing enterprise customer base, Onyx is well-positioned. The challenge now? Scaling up while staying ahead in a market where AI search is evolving fast.

With just 6 employees and 3000 developers, they can cover edge cases for organizations that are 100x their size. 

Beyond simple retrieval, Onyx’s AI-driven research agent ensures that search results are not just relevant but also refined. By evaluating candidate answers for inconsistencies, hallucinations, or missing elements, the system provides responses that enterprises can trust. This validation process is crucial in high-stakes environments where misinformation can lead to costly mistakes.

Will Onyx become the go-to enterprise search solution, or will established players edge it out? That remains to be seen. But like Chris mentioned, every team in the world should have access to their own knowledge enhanced agents.