Onyx
What it is
Onyx connects company knowledge to AI answers for people and agents. It searches indexed sources in parallel and returns relevant documents while enforcing the permissions attached to those sources. Connectors sync data in the background and support hybrid semantic and keyword search; Onyx says it offers more than 50 connectors. The Business plan lists chat and search, major language model access, custom agents, MCP/OpenAPI actions, web search, deep research, code interpreter, and image generation. It is priced at $20 a month with annual billing, at $20 per user per month, and includes 40+ app connectors and Community Support. Onyx provides APIs for messages, agents, model management, actions, and user authentication. It supports deployment in a customer's cloud or on bare metal, including on-premise and air-gapped environments. The site says documents and embeddings can remain in the customer's infrastructure, data is not used for training, and source permissions are checked at query time. Native desktop apps for Windows, macOS, and Linux connect to an existing Onyx deployment. Slack integrations can provide responses and agents; Microsoft Teams and Zendesk are described as coming soon.
For and against
For
- Search respects source permissions at query time.
- Connectors sync data in the background.
- Supports customer-cloud, bare-metal, and air-gapped deployments.
- Desktop apps support Windows, macOS, and Linux.
Against
- Business pricing is $20 a month with annual billing.
- Desktop apps connect to an existing Onyx deployment.
- Microsoft Teams and Zendesk are described as coming soon.
The verdict
Onyx combines enterprise knowledge search with AI tools and multiple deployment choices. Teams should account for annual billing and the existing-deployment requirement for desktop apps.
Who it suits
Onyx is aimed at company teams that want staff and AI agents to search organizational information with source permissions enforced. Its deployment options also suit organizations seeking cloud, on-premise, or air-gapped setups.
Plans and prices
Business
Chat/search · major LLMs · custom agents · MCP/OpenAPI actions
Read 20 Oct 2026 · onyx.app
Community Edition
Core RAG · deep research · custom agents · actions
Read 20 Oct 2026 · github.com
Enterprise
Business features · OIDC/SAML SSO · on-premise · region-specific deployment
Read 20 Oct 2026 · onyx.app
On the record
- Actions and MCP
- Five built-in actions cover internal search, web search, Python code execution, image generation and a beta coding agent; additional actions use OpenAPI or MCP with token or OAuth authentication.docs.onyx.app · read 20 Oct 2026
- API
- Nearly every feature is exposed through a REST API using JSON, with Bearer API keys or personal access tokens and an OpenAPI explorer at /api/docs.docs.onyx.app · read 20 Oct 2026
- API permissions
- API keys act as service accounts whose group memberships control chat, search, agent, connector and administrative access; custom groups are Enterprise-only.docs.onyx.app · read 20 Oct 2026
- Chat
- The chat interface supports file and URL context, model selection, citations, projects, sharing and multiple AI actions.docs.onyx.app · read 20 Oct 2026
- Chrome extension
- The Chrome extension provides new-tab search, a sidebar, selected-text queries and keyboard shortcuts, with full-page questioning listed as future functionality.docs.onyx.app · read 20 Oct 2026
- Company founders
- Crunchbase lists Chris Weaver and Yuhong Sun as founders.crunchbase.com · read 20 Oct 2026
- Company legal name
- Crunchbase identifies the legal entity as DanswerAI Inc., formerly known as DanswerAI and Danswer.crunchbase.com · read 20 Oct 2026
- Company location and size
- Crunchbase places the company in San Francisco, California, United States and lists 11–50 employees.crunchbase.com · read 20 Oct 2026
- Connector behavior
- Connectors sync data in the background and support hybrid semantic and keyword search across connected sources.onyx.app · read 7 Oct 2026
- Connectors
- Onyx keeps connected application data synchronized and can enforce source permissions in Enterprise Edition; documents can also be pushed through API or file upload.docs.onyx.app · read 20 Oct 2026
- Custom agents
- Users can define agents with natural-language instructions, scoped connector or uploaded knowledge and selected actions, then share them with users or groups.docs.onyx.app · read 20 Oct 2026
- Deep research
- Deep Research runs repeated reasoning, research and action cycles, can take several minutes and may consume more than ten times normal inference tokens.docs.onyx.app · read 20 Oct 2026
- Deployment
- Onyx can run as self-hosted Standard or Lite, in Docker Compose, Kubernetes/Helm or cloud infrastructure, and is also offered as managed Onyx Cloud.docs.onyx.app · read 20 Oct 2026
- Desktop apps
- Official documentation provides downloads for Windows EXE, macOS DMG and Linux DEB, with additional MSI, RPM and tarball formats via GitHub Releases.docs.onyx.app · read 20 Oct 2026
- Desktop platforms
- The native desktop app is available for Windows, macOS, and Linux and connects to an existing Onyx deployment.onyx.app · read 7 Oct 2026
- Developer API
- Onyx supports APIs for messages, agents, LLM management, actions, and user authentication.onyx.app · read 7 Oct 2026
- Documentation depth
- The documentation includes feature guides, API authentication and reference guidance, Docker, Helm, Terraform, cloud deployment, security hardening, authentication and release notes.docs.onyx.app · read 20 Oct 2026
- Features
- The Business plan lists chat and search, major LLM access, custom agents, MCP/OpenAPI actions, web search, deep research, code interpreter, and image generation.onyx.app · read 7 Oct 2026
