Complete Context for Your Humans and Agents

Onyx connects to your enterprise knowledge and gives humans and agents accurate, fast, and token-efficient context.

The world's top teams trust Onyx to be their primary interface to AI.
1M+queries per week
1,000+teams
Retrieval Quality

Onyx Finds What the Others Miss

Onyx Searchall sources in parallel
Why did our AWS bill jump $40K this month?
Searching indexed sources
GitHubDriveSlackConfluence
→ 10 relevant docs · permissions enforced
Decision Tear it down · $40,200/mo recovered
Traditional MCP searchone source at a time
Why did our AWS bill jump $40K this month?
search_drive("cost export") · 9 files
search_slack("aws bill") · 0 results
search_confluence("cloud costs") · 74 pages
search_github("perf-test") · rate limited
context window full · 3 of 10 relevant docs
Agent
Decision Missed it · the cluster keeps running
Speed and Cost

Faster Answers at Lower Cost

Onyx's state-of-the-art search pipeline helps you do more with AI while saving time and tokens.

Onyx

Onyx: 6.7 seconds and $0.69 in this example.

Live app search

Live app search: 10 seconds and $1.15 in this example.

Both answers are ready.

Benchmark

State of the Art Search

Onyx sends your question out in several forms to make sure it knows exactly what you are looking for. We've outperformed industry leaders on our open source benchmark.

EnterpriseRAG-Bench · overall500 questions
OnyxGPT-5.4
72.4
OpenClaw
68.2
OpenAI File Search
61.0
Amazon Q (Kendra)
49.0
Azure AI Search
48.4
Vertex AI Search
41.9
Onyx win rate · two LLM judges99 questions
50% EVEN0.0%ONYX WIN RATE
ChatGPT
Onyx
50% EVEN0.0%ONYX WIN RATE
Claude
Onyx
50% EVEN0.0%ONYX WIN RATE
Notion AI
Onyx
Head to Head

Answers You Can Trust

Tested on 99 real workplace questions across 220K internal documents, scored blind by two independent LLM judges. A win means judges preferred the Onyx answer in a head-to-head comparison; 50% is even.

Self-Hosted

Don't Give Up Control of Your Data

Onyx runs entirely in your cloud or on bare metal. The index, the embeddings, and the model traffic all stay inside the boundary your security team already owns, which is why the teams with the strictest requirements can run it at all.

  • Documents and embeddings stay in your infrastructure

  • No training on your data, by anyone, ever

  • Source permissions enforced at query time

  • Open source, so the code is auditable end to end

Your networkyour vpc · 10.0.0.0/16IngestData planeInferenceConnectors40+ sources · permissions syncedOpenSearchhybrid index + vectorsPostgresdocuments · metadataInferencevLLM · self-hostedGPU nodesyour cloud accountegressdenyYour networkyour vpc · 10.0.0.0/16IngestData planeInferenceConnectors40+ sources · permissions syncedOpenSearchhybrid index + vectorsPostgresdocuments · metadataInferencevLLM · self-hostedGPU nodesyour cloud accountegressdeny
Enterprise ready

Security & Compliance

Flexible deployment supporting single tenancy / within VPC / on-premise.

Onyx can be run fully air-gapped with no external third parties.

routing toAnthropic1 click to change

Any Model

Swap providers in one click, route each team to a different model, or run open weights on your own GPUs. When the frontier moves, you move with it.

50+ syncedAWSTestrail SVG Icon

50+ Connectors

Every tool your company already works in, synced continuously and indexed with the permissions each source enforces.

Learn more →

White-Glove Support

A dedicated forward-deployed engineer helps you identify key use cases, configure Onyx, and roll it out to your team.

Interfaces

Use Onyx Anywhere

Use Cases

Improve Productivity

for Every Department

Company Wide

Empower every member with secure access to GenAI and knowledge

Engineering

Ship faster with Generative AI and full context

Sales

Close more deals with instant access to every conversation and product update

Support

Answer questions confidently across your entire product

Onyx AI | Open Source Enterprise Search & AI Assistant