Onyx (formerly Danswer) is an AI chat and search platform that connects to 40+ company applications, builds custom agents, and can run as a hosted product or self-hosted. Pick it when your team needs answers pulled from your own scattered documents and tools, not just a general chatbot. Business is $20/user/month (annual) with a free trial; full self-hosting needs your own server infrastructure for indexing, workers and blob storage.
Other good fits
Building custom AI agents scoped to your organisation's own knowledge and tools.
+ Pros
Read pros
Onyx provides a chat and search interface over your organisation's own knowledge, connecting to 40+ applications so people can ask a question and get an answer pulled from the right document, wiki page or ticket instead of manually searching each tool. It also supports building custom AI agents, deep research across connected sources, image generation, and access to major LLM providers, plus a developer API for building on top of it.
Pick it when…
Your team's knowledge is scattered across many tools and people waste time searching each one separately.
You want to build custom agents scoped to your organisation's actual documents and workflows.
You need chat and search in one interface with Slack integration and a code interpreter for technical teams.
− Cons
Read cons
A connected-knowledge chat tool is only as good as its indexing and source access controls; check who can see what before connecting sensitive sources, since a badly scoped connection can expose more than intended. Self-hosting the full platform (not just a lightweight demo) is a real infrastructure commitment: indexing workers and blob storage don't run themselves.
Pick something else when…
You just need a general chatbot with no connection to internal data.ChatGPT or Claude cover that without any indexing setup.
You want a full team AI workspace with reusable instruction sets, not primarily search.Langdock is built around that different emphasis.
You want to prototype a retrieval pipeline yourself rather than use a packaged product.LangChain gives you that code-level control instead.
Avoid if
You want the free tier without operating server infrastructure; full self-hosted Onyx needs indexing workers and blob storage you manage.
Workflow
Connect your organisation's applications (docs, wikis, ticketing, chat) → let Onyx index the content → ask a question in the chat interface → review the answer and its cited sources → refine the connected sources or agent configuration if answers are missing context. Someone still needs to own which sources are connected and check answers against source documents periodically.
Sign-up available; feature limits not detailed on the pricing page
Business
$20/user/month (annual billing)
Chat, search, custom agents, web search, code interpreter, Slack integration
Enterprise
Custom
OIDC/SAML SSO, on-premise and region-specific deployment, white-labeling, dedicated support
Full self-hosted Onyx (rather than the hosted Business/Enterprise plans) needs your own server infrastructure for indexing, background workers and blob storage; that's real operational work, not a one-container deployment. Running effort on any plan includes reviewing which sources are connected, since a connected source becomes searchable by anyone with access to the corresponding agent or chat.
Restrictions
Full Onyx (self-hosted) needs server infrastructure for indexing, background workers and blob storage; a lightweight local trial doesn't represent the operational footprint of a production deployment. Check current documentation for exact infrastructure requirements before committing to self-hosting at scale.
First project
Documentation-based suggestion — not run here.
Connect one source and test a real question.
Start a free trial and connect just one application your team actually uses (a wiki or a docs tool).
Let indexing complete.
Ask a question you already know the answer to, so you can verify accuracy.
Check the cited source against the actual document to confirm it's not paraphrasing incorrectly.
Done when you've verified at least one answer traces correctly back to its cited source, confirming the connection and indexing actually work before connecting more sensitive sources.
Links & evidence
Documentation-based · not hands-on · Sep 23, 2026
Rests on Onyx's pricing page and homepage, checked 23 September 2026. No account was created and no source was connected for this review.
Onyx pricing · Sep 23, 2026 — Free trial, Business $20/user/month (annual), Enterprise custom; feature breakdown per tier.
Search connected applications with source-cited agentic retrieval.
Create agents with instructions, knowledge, and actions.
Run hosted or self-hosted with Docker or Kubernetes.
Onyx (onyx-dot-app) · free-tier · self-host
enterprise search
team chat
agentic rag
app connectors
self-hosted docker
kubernetes deployment
role-based access
custom ai agents
code interpreter
search-cite
talk-think
draft-docs
open-source
View sourced imageExisting Connectors administration view from the official Onyx documentation. The connector counts are the publisher's example configuration.
Onyx documentation — Admin panel overview
Owner-approved use of public official interface screenshots, as recorded in public-visuals-batch-1.json, 13 September 2026. Copyright retained by Onyx and its contributors; attribution retained. Editorial product illustration; no open licence asserted. Converted on 15 October 2026 to a 1600px-wide WebP (quality 80) for faster page loads; the original image remains in the media library.
