Onyx raises $10M to help teams find knowledge faster with open source AI agents
We started Onyx (formerly known as Danswer) nearly one year ago to make information more accessible at work. We were inspired by our experience working at companies big and small, where there was so much institutional knowledge buried in disparate knowledge bases, shared drives, and individual files. Onyx is the open source solution companies use to solve this challenge.
Today, we’re proud to be working with dozens of amazing enterprise customers. From household names like Netflix (where we’re deployed to all 14,000+ employees) to extremely security conscious defense companies like Thales Group, organizations of all sizes are choosing Onyx to help their teams work smarter.
We’re also excited to announce our $10M seed round co-led by Khosla Ventures and First Round Capital, with participation from Y Combinator and amazing angels like Gokul Rajaram (Coinbase, Pinterest), Arash Ferdosi (Dropbox), Amit Agarwal (Datadog), and others.
Modern teams aren’t making use of their massive knowledge bases
As teams scale, knowledge becomes scattered (and essentially lost) across a sprawl of SaaS tools. We’ve all been there: frantically scanning through Slack, Confluence, Jira, and Google Drive just minutes before a meeting or while on-call trying to resolve an outage — only to be let down by outdated keyword-based searches. This leads to critical downstream impacts:
- Higher churn and lost opportunities because of slow responses to customers and prospects.
- Bad decisions due to partial or incomplete information.
Wasted time and resources from duplicated work and team members unable to unblock themselves.
Onyx acts as your most knowledgeable coworker, available 24/7
Onyx indexes knowledge across more than 40 tools such as Google Drive, Salesforce, Sharepoint, GitHub, and more. Instead of pinging someone over Slack/Teams and waiting minutes or hours to get a response, you can ask Onyx anything and get an answer instantly.
Traditional search can't solve this problem — you need an AI research agent
Many questions at work can’t be answered by a simple query. Onyx is the first AI research agent designed for the workplace that searches like a human would. Our agent explores parallel pathways and iteratively digs down promising directions to retrieve the most complete context for each question.
For example, if someone in sales wanted to know, “How did we close Bitwarden so quickly?” Onyx’s AI agent would first figure out what it needs to answer this question: details of the Bitwarden sales process AND an understanding of a standard sales process. The agent would run parallel searches through Salesforce, Gong, email, and Slack to get answers to both of these, and then combine them to build a coherent comparison. If the agent finds that a Proof of Concept was run, it will adjust mid-flight and research to see what the Bitwarden PoC process was like and compare it with others.
After generating a candidate answer, Onyx then evaluates it to see if there are contradictions, hallucinations, or incomplete aspects. Finally, the agent runs a set of refinement steps and additional searches to fill in the missing pieces and correct any issues with the preliminary answer.
Our AI research agent improves both search quality at scale as well as the completeness and reliability of the generated answer.
The future of AI at work is open source
Large language models and the revolution in deep learning have not only changed the foundational technology used for enterprise search — it’s also made open source the best approach for a few reasons:.
- Your data stays secure. Every team should be able to use generative AI without needing to hand over control of their data. We built Onyx to be the most configurable and easiest to self-host. All data processing happens within the Onyx deployment and is stored locally. You choose the LLM provider or even plug in local LLMs for a truly airgapped system. We also offer a hosted solution (Onyx Cloud) which runs on AWS and uses minimal third-party services.
- Complete customization. You want your AI assistants to do more than give answers. Leveraging the extensibility of the OSS nature of Onyx, the University of California, San Diego (where we’re live with 37,000 users) has built human-in-the-loop agent workflows to automate common tasks. Ramp (where Onyx not only powers internal knowledge but parts of the Ramp product) has built multi-step AI agents to auto-resolve >93% of actionable customer service requests. As AI quality gets better and the cost of development gets cheaper, open source allows you to craft the most suitable custom experiences for your needs.
- Ship Faster. With the help of our developer community of more than 3,000 strong, we can cover all your edge cases as if we were an organization 100x our size. 125 unique individuals have contributed code back to Onyx and more than half of our integrations have been built by contributors. We even have contributions from numerous world class teams like Zendesk, Volkswagen, Harvard University and many others.
Unlocking knowledge for teams of all sizes
In the future, every team in the world should have access to their own knowledge enhanced agents. At Onyx, we’re making this technology incredibly accessible and easy to get started.
To try out Onyx cloud for free — visit our signup page to get started in <1 minute.
If you’d like to self-host Onyx, check out our GitHub and Quickstart Guide.
Or book a demo with us to learn more about the Onyx Enterprise Edition.