Onyx is an open-source AI chat and enterprise search platform that connects to your company’s internal tools, documents, and data sources to give every employee instant, cited answers grounded in your organisation’s actual knowledge. It solves the problem of fragmented company knowledge — the daily frustration of searching through Slack, Confluence, Google Drive, Jira, and Salesforce only to find outdated or irrelevant results.

It is best suited for engineering teams, IT administrators, support teams, and knowledge-heavy organisations that need a secure, self-hostable AI assistant without sending sensitive data to third parties.

Its core AI capabilities include hybrid retrieval-augmented generation (RAG), agentic deep research, permission-aware search, custom AI agents, MCP-powered actions, and a six-stage retrieval pipeline that dramatically reduces hallucinations.

What sets Onyx apart from alternatives like Glean and Microsoft Copilot is its open-source foundation — companies can self-host it on any infrastructure, inspect the code, choose any LLM, and fully customise it without vendor lock-in.

Close alternatives: Glean, Microsoft Copilot, Guru, Notion AI.

Onyx Key Features

Permission-Aware Enterprise Search: Connects to 40+ company tools and automatically inherits document-level access controls from each source — so employees only see answers from content they are authorised to access, eliminating data leakage risks.
Six-Stage Agentic RAG Pipeline: Runs a multi-step retrieval process including parallel query generation, hybrid vector and keyword search, LLM-based chunk selection, context expansion, and grounded answer synthesis with inline citations — delivering far more accurate responses than single-pass RAG systems.
Deep Research Mode: Breaks complex workplace questions into sub-questions, explores multiple research paths in parallel across all connected knowledge sources, and synthesises a comprehensive cited report — ranked top of leaderboard as of February 2026. [Source: GitHub]
Custom AI Agents: Lets teams build specialised AI assistants with unique instructions, dedicated knowledge sets, and defined action capabilities — enabling purpose-built agents for sales, support, engineering, HR, or any other department.
MCP Actions and External Integrations: Allows Onyx agents to interact with external systems via Model Context Protocol (MCP) and OpenAPI, going beyond search to take actions in connected tools on behalf of users.
Self-Hosted and Cloud Deployment Options: Deploy Onyx on-premises, in a VPC, in a fully air-gapped environment, or on Onyx Cloud — giving organisations complete control over where their data lives and how it is processed.
LLM Flexibility with No Model Lock-In: Works with all major proprietary LLMs (OpenAI, Anthropic, Google Gemini) and self-hosted open-source models (Ollama, vLLM, LiteLLM) — allowing teams to swap models without rebuilding their knowledge layer.
OnyxBot for Slack: Embeds an AI assistant directly inside Slack that automatically answers team questions and customer inquiries using the company’s internal knowledge, without employees needing to open a separate app.
Real-Time Connector Syncing: All 40+ connectors sync knowledge updates in real time, so answers always reflect the latest version of connected documents, tickets, and conversations — not a stale snapshot.
Granular RBAC and SSO: Supports role-based access control, Google OAuth, OIDC and SAML SSO, SCIM user provisioning, and query history audit logs — meeting enterprise security and compliance requirements out of the box.

Onyx Key Customers

Customer	What They Do	What Was Achieved Using Onyx	Source
Ramp	Fast-growing fintech / expense management platform	Onyx answers thousands of questions per week at Ramp. Tony Rios, Director of Product Ops, stated it delivered a 30x ROI and was the only AI tool with reliable enough answers to drive productivity at scale. Ramp also built multi-step AI agents on Onyx to auto-resolve over 93% of actionable customer service requests.	Ramp Case Study — Official Onyx case study
Netflix	Global streaming and entertainment company	Onyx is deployed across all 14,000+ Netflix employees as an internal knowledge platform.	Onyx Seed Round Blog — Official Onyx announcement
Thales Group	French aerospace and defence multinational	Adopted Onyx for its security-critical internal knowledge search, choosing it specifically because of the self-hosted deployment model that keeps data on-premises.	Onyx Seed Round Blog — Official Onyx announcement
University of California, San Diego (UCSD)	Public research university	Deployed Onyx for 37,000 users and built human-in-the-loop agent workflows on top of it to automate common administrative tasks.	Onyx Seed Round Blog — Official Onyx announcement
Bitwarden	Open-source password manager and cybersecurity company	Uses Onyx for internal knowledge management, cited as a customer in the $10M seed round announcement.	GlobeNewswire Press Release

Who is the Founder / CEO of Onyx?

Yuhong Sun is the Co-founder and Co-CEO of Onyx. He specialises in machine learning and natural language processing, and previously worked at Alation, Inc., where he developed deep learning systems for NL-to-SQL translation and semantic enterprise search. He studied at the University of California, San Diego. The idea for Onyx came directly from his frustration with how institutional knowledge was buried and unsearchable at growing tech companies.

Chris Weaver is the Co-founder and Co-CEO of Onyx. He previously served as a Technical Lead on the anti-fraud engine team at Robinhood, building systems to detect and prevent fraudulent activity at scale. He also studied at the University of California, San Diego, and is notably a former top-100 North American ranked League of Legends player. The two co-founders previously built two other companies together before starting Onyx — a Twitch stats tracker and a speciality keyboard comparison site — before arriving at Onyx (originally called Danswer) in 2023.

Where is Onyx Headquartered?

Onyx is headquartered in San Francisco, California, USA. The company was founded in 2023 and operates under the legal entity name Onyx (previously known as Danswer, a portmanteau of “deep answer”).

Onyx News and Funding

In March 2025, Onyx raised a $10 million seed round co-led by Khosla Ventures and First Round Capital, with participation from Y Combinator and prominent angel investors including Gokul Rajaram (Coinbase, Pinterest), Arash Ferdosi (Dropbox co-founder), Amit Agarwal (President, Datadog), Immad Akhund (Mercury co-founder), and Kulveer Taggar (Zeus Living co-founder), among others. [Source: TechCrunch]

Onyx is also supported by YCombinator – Onyx Open Source AI Chat on YC

At the time of the funding announcement, Onyx’s open-source project had been deployed by thousands of organisations and was processing hundreds of thousands of questions every week. The GitHub repository has accumulated over 20,000 stars and more than 3,000 developer contributors. The product was originally launched as Danswer in 2023 and rebranded to Onyx ahead of the seed round. [Source: Onyx Blog]

Who Should Use Onyx?

Here are some of the best use cases for Onyx:

Giving every employee instant, cited answers about internal processes, product details, and company history without pinging a colleague on Slack.
Enabling engineering teams to search across code repositories, Jira tickets, Confluence pages, and Slack threads in one query.
Building a privacy-first AI assistant for organisations in regulated industries (defence, healthcare, finance) that cannot send data to external cloud providers.
Accelerating new employee onboarding by making institutional knowledge searchable and accessible from day one.
Automating high-volume customer support queries using internal product knowledge, as Ramp did to auto-resolve 93%+ of actionable requests.

This makes it ideal for the following customer profiles:

Engineering and Product Teams who need to search across codebases, documentation, tickets, and communication channels in one place without context-switching.
IT Administrators and Platform Engineers looking to deploy a self-hostable, permission-aware AI knowledge layer on top of their existing tool stack.
Customer Support and Success Teams wanting to reduce response times by giving agents instant access to every product update, policy document, and past conversation.
Sales Teams needing real-time access to competitive intelligence, product specs, and past deal notes to close faster.
Security-Sensitive Organisations in aerospace, defence, healthcare, and financial services that require on-premises or VPC deployment with no data leaving their infrastructure.
Developers and Open-Source Enthusiasts who want to self-host, customise, or contribute to an enterprise AI platform without vendor restrictions.
Universities and Research Institutions wanting to deploy a large-scale knowledge assistant for students and staff, as UCSD did with 37,000 users.

Onyx Pros

Superior Answer Accuracy vs. Competitors — Onyx’s own benchmarks tested on 99 real workplace questions across 220,000 internal documents show a 64% win rate against ChatGPT, 68.1% against Claude, and 76% against Notion AI when scored by independent LLM judges. While self-reported, the methodology is documented and the results have been cited by enterprise customers as validation of real-world accuracy. [Source: onyx.app/answers]
Open-Source Transparency Drives Enterprise Trust — TechCrunch noted that Onyx’s open-source approach eliminates the lengthy enterprise sales cycle typical of closed competitors like Glean, allowing companies to experiment, self-host, and verify the code before committing. Multiple enterprise customers explicitly cited transparency as a key reason for choosing Onyx over alternatives. [Source: TechCrunch]
30-Minute Setup to Production-Ready Deployment — Companies can get Onyx running in approximately 30 minutes using Docker Compose, making it unusually fast to deploy for a platform of this depth.
Explosive Organic Growth Driven by Community — With 20,000+ GitHub stars, 3,000+ code contributors, and over half of all integrations built by the open-source community, Onyx has a development velocity that far exceeds what its team size alone could produce. This community compounds both the connector coverage and the reliability of the platform. [Source: Onyx Blog]
Validated by High-Scale Enterprise Deployments — Being deployed to all 14,000+ Netflix employees and 37,000 UCSD users provides meaningful proof that the platform scales reliably under real enterprise conditions. [Source: Onyx Blog]
No Model Lock-In — A Medium reviewer who tested Onyx for enterprise RAG deployments highlighted that the ability to swap LLM providers — from OpenAI to Anthropic to self-hosted Ollama — without rebuilding the knowledge layer is a significant differentiator from proprietary alternatives. [Source: Medium — AI Rabbit]

Onyx Cons

Self-Hosting Requires Meaningful DevOps Overhead — Onyx’s own documentation acknowledges that self-hosting “requires more operational overhead compared to the Cloud offering” and recommends the cloud version unless self-hosting is specifically necessary. Teams without dedicated DevOps resources may find setup and maintenance challenging, particularly at scale. [Source: Onyx Documentation]
Community Edition Has No Dedicated Support — The free self-hosted Community Edition comes with community-only support via Discord. Organisations that need SLAs, dedicated engineers, or priority response must upgrade to the Business ($20/user/month) or Enterprise tier. This can be a meaningful gap for teams that hit issues in production during off-hours. [Source: onyx.app/pricing]
Connectors for Self-Hosted Open-Source Tools Are Limited — A GitHub issue from the community noted the absence of connectors for popular self-hosted alternatives like Gitea, MkDocs, and Vikunja — tools commonly used by privacy-focused teams who are also likely to self-host Onyx. This is an ironic gap given the target audience. [Source: GitHub Issue #776]
Competitive Field with Well-Funded Rivals — TechCrunch noted that Onyx competes with Glean, which has raised over $600 million in venture funding, giving it significantly more resources for enterprise sales, integrations, and feature development. Onyx’s open-source advantage helps, but the sales motion at the largest enterprises remains challenging. [Source: TechCrunch]
Benchmarks Are Self-Reported — The accuracy win-rate figures published on the Onyx website (64% vs ChatGPT, 76% vs Notion AI) are based on Onyx’s own benchmark methodology using two independent LLM judges. While the methodology is documented, independent third-party verification of these figures has not been published, which means buyers should treat them as directional rather than definitive. [Source: onyx.app/answers]
No Published G2 or Capterra Reviews — As of April 2026, Onyx does not have a significant volume of verified user reviews on major third-party review platforms like G2 or Capterra. This makes independent comparison with other enterprise tools more difficult for buyers relying on peer review platforms during procurement.

Onyx Integrations

Onyx connects to 40+ enterprise tools out of the box via plug-and-play connectors that sync in real time and automatically inherit permission settings from each source. Key connectors include:

Communication: Slack, Microsoft Teams, Discord, Email (Gmail, Outlook)
Documentation and Knowledge: Confluence, Notion, GitBook, SharePoint, Google Drive, Dropbox, OneDrive
Code and Engineering: GitHub, GitLab, Jira, Linear, Asana
CRM and Sales: Salesforce, HubSpot, Gong
Support: Zendesk, Intercom, ServiceNow, Front
Identity and HR: Rippling, Okta (via SSO/SCIM provisioning)
Data and Analytics: Airtable, Looker, Guru

Beyond native connectors, Onyx supports custom connectors via its open API and MCP protocol for proprietary internal systems.

The full connector list is available at docs.onyx.app/admins/connectors/overview.

Onyx Free Plan

Onyx’s free option is the Community Edition (CE), available under the MIT open-source licence. It is a permanent free tier for teams willing to self-host.

The Community Edition includes the full core feature set: AI chat with RAG, enterprise search, custom AI agents, MCP actions, 40+ connectors, web search, deep research, code interpreter, image generation, voice mode, RBAC, Google OAuth, query history, and multi-user collaboration.

Key limitations of the Onyx Community Edition worth noting:

Requires self-hosting (Docker, Kubernetes, or cloud VM) — Onyx Cloud is the paid managed option.
No OIDC/SAML SSO (Enterprise Edition only).
No on-premises or region-specific managed deployments.
No white-labelling.
No custom integrations support from the Onyx team.
No data exports.
No dedicated support — community support via Discord only.
No enterprise SLA.

Why you might need an Onyx paid plan:

If you want Onyx managed in the cloud without DevOps overhead, the Business plan at $20/user/month provides Onyx Cloud access with all core features.
If you need SSO, on-premises managed deployments, white-labelling, compliance features, invoice billing, or dedicated support, the Enterprise plan is required.
If your team has compliance or security requirements that need audit trails and region-specific data residency.

Onyx Paid Plan Pricing

Plan	Price	Key Features	Who It’s Suitable For
Community Edition (Free)	Free, self-hosted forever	Full AI chat, RAG, agents, MCP, 40+ connectors, deep research, code interpreter, RBAC, Google OAuth, community support via Discord.	Developers, startups, and technically capable teams who want full control and can manage their own hosting.
Business	$20/user/month (annual billing)	Everything in Community Edition, plus: Onyx Cloud managed hosting; access to all major LLMs; custom agents; Slack integration; developer APIs; usage dashboards; basic auth; RBAC; encryption of secrets; community support.	Teams of any size that want the full Onyx feature set on managed cloud without self-hosting.
Enterprise	Contact for pricing	Everything in Business, plus: OIDC/SAML SSO; on-premises and region-specific managed deployments; early access to features; white-labelling; custom integrations; data exports; invoice billing; volume discounts; higher education pricing; dedicated support; enterprise SLA.	Large enterprises, defence and regulated industry organisations, and institutions needing compliance, on-premises deployment, or dedicated support.

Pricing verified from onyx.app/pricing as of April 2026.

Onyx Discounts

Community Edition is Permanently Free: The self-hosted MIT-licenced version has no cost, no trial expiry, and no feature restrictions on core capabilities — making it the most accessible free tier in the enterprise search category.
Higher Education Pricing: The Enterprise plan explicitly lists higher education pricing as an available option, meaning universities and research institutions can negotiate a discounted rate. Onyx should be contacted directly for specifics.
Volume Discounts: Enterprise plan includes volume-based discounts for large user counts, negotiated directly with the sales team.
No Published Annual vs Monthly Discount: The Business plan is listed at $20/user/month on annual billing. A monthly billing option and its rate are not published on the pricing page.
No NGO or Startup Discount Publicly Listed: Onyx does not currently advertise special pricing for nonprofits or early-stage startups beyond the free Community Edition.

Onyx Alternatives

Alternative	Strengths	Limitations
Glean	100+ connectors; deep personalisation; strong enterprise sales motion; AI agents and knowledge graph	Not open source; cloud-only; estimated ~$50/user/month; no self-hosted option; lengthy enterprise sales cycle
Microsoft Copilot	Deep Microsoft 365 integration; familiar UI; included in some M365 plans	Locked to Microsoft ecosystem; no LLM flexibility; limited connector coverage outside Microsoft tools
Guru	Strong knowledge base management and verification workflows; easy for non-technical teams	Less powerful RAG architecture; focused on curated knowledge rather than raw enterprise search
Notion AI	Excellent for teams already in Notion; good writing and summarisation features	Only searches within Notion; no cross-tool enterprise search; benchmark data shows Onyx outperforms it on workplace Q&A
Open WebUI	Feature-rich open-source chat UI; very easy to self-host; large community	No enterprise connectors; no permission-aware search; no knowledge indexing — just a chat interface
Confluence + AI	Already where many teams store docs; built-in search improvements	Fragmented; only searches Confluence content; not a cross-tool enterprise search solution

How is Onyx Different from Other Alternatives?

Onyx’s core differentiator is the combination of open-source transparency, self-hostability, LLM flexibility, and a genuinely advanced RAG pipeline — all at a price point that undercuts closed alternatives significantly. While Glean is the market leader with the broadest connectors and enterprise sales presence, it costs an estimated ~$50/user/month and requires cloud-only deployment. Onyx gives teams 80% of Glean’s capability at $20/user/month on cloud, or entirely free with self-hosting.

For security-conscious organisations in regulated industries — aerospace, defence, healthcare, government — the ability to run Onyx in a fully air-gapped on-premises environment is a capability that most competitors cannot match at any price.

Choose Onyx when data control, LLM flexibility, and cost transparency matter as much as the quality of the answers themselves.

AI Enterprise Search

Onyx: Best Open-Source AI Enterprise Search Claude Alternative