Onyx is a self-hostable, MIT-licensed AI search and chat platform with 50+ connectors. It does what Glean does, without the $50/user/month price tag or the six-figure minimum contract.

Onyx is a self-hostable, open source AI search and chat platform that replaces what Glean sells for a minimum of roughly $60,000 a year, and it ships as a free community edition under the MIT license.

That sentence deserves a moment. Glean is one of the most heavily marketed enterprise AI search tools on the market. It connects your Slack, your Google Drive, your Confluence, your Salesforce, and a dozen other tools into a single searchable AI layer that can answer questions, surface documents, and run research tasks. It is genuinely useful. It is also priced to reflect that: industry pricing analyses put Glean's base cost at $50 or more per user per month, with a minimum contract floor of roughly 100 seats, putting the entry point at around $60,000 annually before you add AI add-ons, support fees, or infrastructure. For mid-to-large organizations, total annual spend is regularly reported in the $350,000 to $480,000 range.

Onyx does the same job. Fewer enterprise bells, but the same core promise: your team asks a question in plain English, and the system searches across every tool you have connected and returns a sourced, AI-synthesized answer.

What Onyx Actually Does

Think of Onyx as a chat interface with a long memory of every document, message, and database record your company has ever produced, pulled live from wherever those things live.

Out of the box, Onyx ships with more than 50 connectors: Slack, Google Drive, Confluence, Notion, Jira, GitHub, Salesforce, HubSpot, Zendesk, SharePoint, Gmail, and more. When someone on your team asks "what did we decide about the refund policy last quarter," Onyx retrieves the relevant Slack threads, Notion pages, or email threads, synthesizes them through a language model of your choice, and returns an answer with citations so the reader can verify the source.

Beyond basic search, Onyx supports custom AI agents, deep research flows, code execution in a sandbox, image generation, and a voice interface. It ranked at the top of an open deep research leaderboard as of February 2026, which is a meaningful signal for teams that want more than a glorified keyword search.

The most recent release, v4.1.3, shipped June 17, 2026. The project is actively maintained with over 8,600 commits and a visible CI pipeline.

The Setup Reality

Here is where honesty is required. Onyx is not a SaaS you log into on Monday afternoon. It is software you run.

The install is genuinely straightforward for a team with any technical capacity. The single-command path is real:

curl -fsSL https://onyx.app/install_onyx.sh | bash

That gets you running in Docker on a single machine. For teams that need production-grade reliability, Onyx supports Kubernetes and Helm deployments with guides for AWS, GCP, and Azure. A standard deployment runs several containers, a vector and keyword index, a Redis cache, and background workers for syncing connectors. You will need a server or cloud instance with enough memory to run all of that, probably a $100 to $300 per month cloud instance depending on team size and connector volume. That is your infrastructure cost.

You also need someone who can perform occasional maintenance when updates drop, connector credentials rotate, or the indexing job needs a restart. That is not a full-time role for a team of 50, but it is not zero effort either. If your company has no one who can run a Docker stack, Onyx Cloud exists at $20 per user per month, which is still less than half of Glean's per-seat rate.

The community edition (CE) covers all the core functionality: chat, RAG search, agents, and connector integrations. The enterprise edition adds SSO via SAML and OIDC, role-based access controls, usage analytics, query history auditing, and white-labeling. For most teams evaluating whether to replace or skip Glean, the community edition will be sufficient.

What You Give Up

A few honest gaps. Glean has invested heavily in its connector depth and the behavioral intelligence it layers on top of raw search, including surfacing documents by who in your org actually uses them and how recently. Onyx's search quality is strong but approaches the problem differently, through hybrid vector and keyword indexing rather than usage-graph signals.

Glean also comes with an enterprise support tier, a customer success team, and contract-backed SLAs. Onyx's community support is a Discord server. The enterprise edition adds formal support, but the company is smaller than Glean and the resources reflect that.

For regulated industries or companies with strict compliance requirements, the self-hosted path is actually an advantage: your data never leaves your infrastructure. But standing that up to the satisfaction of a security team takes real work, and Onyx does not do that work for you.

The Math for a 50-Person Team

Glean at 50 seats and $50 per user per month is $2,500 per month, or $30,000 per year, and that likely sits below the minimum contract floor, meaning most buyers would be paying for more seats than they need or negotiating a custom arrangement.

Onyx self-hosted at 50 users costs whatever your cloud server runs, realistically $100 to $200 per month for a properly sized instance, plus the occasional hour of a technical employee's time for maintenance. At $150 per month in infrastructure, that is $1,800 per year versus $30,000 or more.

Onyx Cloud at $20 per user per month for 50 users is $1,000 per month, or $12,000 per year. Still less than half of Glean's per-seat rate, with no minimum-seat floor.

Who Should Pay Attention

Onyx makes the most sense for companies that already have internal technical capacity, care about data privacy, and are looking at Glean's pricing and questioning the value. It also makes sense for teams that have been told they are "too small" for Glean's minimums.

The project has 30,400 stars on GitHub, is in production at Netflix, Ramp, and UC San Diego, and shipped its most recent release five days ago. It is not a prototype. It is a production tool that a serious organization can deploy.

The more interesting observation is structural. Glean raised over a billion dollars building a product that Onyx now ships for free. That is not an indictment of Glean, whose distribution and support are real advantages. It is just what happens when the infrastructure for building enterprise software gets cheap enough that a well-resourced open source team can close most of the gap. Buyers who understand that gap, and who have the capacity to run their own stack, are holding real leverage the next time an enterprise sales rep quotes them a six-figure annual contract.