Onyx

Onyx is an open-source enterprise search and AI assistant that connects to a company's apps and answers questions from internal knowledge for teams that want control over their data and models. It is often described as an open alternative to commercial workplace search products. You can run it in your own cloud or data center with any language model, or use the hosted version.

What is Onyx used for?

Onyx is used to give employees one place to find and ask about information spread across many tools.

• Searching across chat, docs, wikis, tickets and code through more than 40 connectors at once
• Enforcing each source's permissions at query time, so people only see what they already have access to
• Chatting with company knowledge, running deep research and using web search and a code interpreter
• Building custom agents for teams such as support, sales or engineering, with actions through MCP or OpenAPI
• Using Onyx from a web app, desktop app, Slackbot, Chrome extension or as an MCP server for other AI tools
• Choosing or switching language model providers, or running open-weight models on your own GPUs

Who Onyx is best for

Onyx suits engineering-led companies, security-conscious organizations, universities and teams in regulated industries that want enterprise search without sending data to a closed vendor. The open-source edition is a good fit for teams with DevOps capacity. Small companies without anyone to run infrastructure may prefer the hosted plan or a fully managed alternative.

Onyx pricing

As of October 2026, the Onyx Community Edition is free and open source under the MIT license, covering core search, AI chat, agents and actions for self-hosting. The hosted Business plan costs $20 per user per month billed annually, with a free trial, and includes chat and search, major LLMs, custom agents, 40+ connectors, deep research, Slack integration, developer APIs and role-based access. Enterprise pricing comes from sales and adds OIDC and SAML single sign-on, on-premise and region-specific deployment, white-labeling, custom integrations and an SLA. Onyx has announced that it is joining Zoom, so watch for changes to plans.

Onyx limitations

• The full self-hosted stack includes indexing, workers and model servers, which takes effort to run
• Some admin features, such as SSO and SCIM, are part of the enterprise editions
• Answer quality depends on how well connectors are configured and kept in sync
• The pending move to Zoom adds some uncertainty about the long-term roadmap

Onyx vs alternatives

Glean is the best-known commercial enterprise search product and suits large companies that want a fully managed service with deep integrations. Guru focuses on verified knowledge cards and an AI answer layer for support and sales teams. Dust lets companies build AI agents on top of their data with a managed service. Atlassian Rovo is the natural pick for teams centered on Jira and Confluence. Onyx wins when open source, self-hosting and model choice matter most.

Tips for getting started with Onyx

Try the hosted trial or the one-line installer on a test server before planning a production rollout. Connect two or three high-value sources first, such as your wiki, Slack and ticketing system, and check permissions with test accounts. Pick a default model per team based on cost and data rules. Add Onyx to Slack early, since that is where many employees will ask questions.

Onyx FAQ

Is Onyx free?

Yes, the Onyx Community Edition is free and open source under the MIT license. The hosted Business plan costs $20 per user per month billed annually.

Is Onyx open source?

Yes. Onyx publishes its code on GitHub, and the Community Edition can be self-hosted with Docker, Kubernetes or Helm.

Can Onyx use my own language model?

Yes, Onyx works with any LLM provider and can run open-weight models on your own GPUs.

Onyx vs Glean: which is better?

Glean suits large companies that want a fully managed enterprise search service, while Onyx is better for teams that want open source, self-hosting and model choice.