Danswer
Onyx (formerly Danswer) · Coding · Leaning · Chatbot
Danswer, now rebranded as Onyx, is an open source AI search and chat platform that connects to your work apps, indexes internal documents, and answers questions with cited sources.
About Danswer
What Is Danswer
Danswer is an open source enterprise search platform that also works as an AI assistant for work. It plugs into the apps a team already uses, pulls their content into a single index, and answers plain language questions with quotes and links back to the original material. The company renamed the product to Onyx in 2024, so the docs, the GitHub repo, and current releases all live under that name.
What makes it useful is that the answers are checkable. A general chatbot guesses from training data. Danswer retrieves passages from your Confluence pages, Slack threads, Google Drive files, and tickets first, then hands those passages to a language model to summarize. Every reply shows where it came from, which is the difference between a demo and something a support team can actually use.
The cost of that control is setup. Self-hosted AI search means someone has to run Docker or Kubernetes, wire up the data sources, pick a model provider, and keep the index from going stale. None of that is hard for an engineer and all of it is a wall for everyone else. Teams that want the capability without the maintenance can use the hosted version instead.
Getting Started
- Deploy the stack. The quickest route is a single docker compose command on a laptop or a server. Kubernetes manifests are published for bigger installs, and Onyx Cloud offers a hosted trial without a credit card.
- Connect your sources. Sign into the admin panel, pick connectors such as Slack, Google Drive, Confluence, Notion, or Jira, authorize them, and choose which spaces or channels get indexed.
- Pick a model. Point Danswer at OpenAI, Anthropic, Gemini, or a local model through Ollama, then paste the API key or endpoint.
- Ask questions. Use internal search when you want a list of matching documents, or the chat view when you want a written answer with citations attached.
- Roll it out. Invite users, assign admin or basic roles, and add the Slack integration so people can ask without opening a new tab.
Product Information
A quick look at Danswer's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- IT and platform teams at mid-size companies
- Support and customer success teams
- Engineering organizations
Tasks
- Answering repeated HR and policy questions
- Tracking down a document you half remember
- Onboarding new hires
- Building a first-pass answer with sources
Scenarios
- A company whose knowledge is spread across Drive, Slack, and Confluence
- A regulated team that can't send internal documents to a vendor cloud
- A 50 to 500 person company that wants Glean-style search without enterprise pricing
Key features
Hybrid Search Across 40+ Connectors
Danswer indexes sources with both keyword and vector search, then blends the two results. Exact matches matter when someone searches for an error code or a ticket number, and meaning-based matches catch the case where the phrasing differs but the topic is the same. Official connectors cover Slack, Google Drive, Confluence, Notion, Jira, GitHub, Salesforce, SharePoint, and dozens more, so company knowledge search usually starts with the tools you already pay for.
Chat That Cites Its Sources
Every answer arrives with quoted passages and links to the documents behind it. You can open the source, read the surrounding text, and decide whether the model summarized it fairly. That citation trail is why the tool holds up for compliance questions, where a confident but wrong answer costs more than no answer at all. It also makes the AI chat over documents feel less like a black box.
Custom Assistants With Instructions
You can build separate assistants, each with its own prompt, knowledge scope, and tone. A support assistant can be told to stay inside the help center and answer in short steps, while an engineering assistant searches design docs and code discussions. Because the scope is limited, the answers stay on topic and the model has less room to invent details.
Agents With Actions and MCP Plugins
Newer releases let assistants take actions, not just answer. You can attach MCP servers or OpenAPI plugins so an agent can file a ticket, look up an order, or query an internal service mid conversation. This is the part that turns internal search into something closer to a workflow tool, and it's also the part that needs the most care around permissions.
Self-Hosted Deployment and Permission Inheritance
The full stack runs in your own cloud, on bare metal, or air-gapped. Documents and embeddings stay in your infrastructure, and the project says it doesn't train on customer data. Enterprise adds permission sync, which mirrors the access rules from the source app at query time, so a user can't read a document in Danswer that they couldn't open in Drive.
Web Search, Deep Research, and Code Interpreter
Beyond internal documents, assistants can search the web, run multi-step research that reads and compares pages, and execute Python for calculations or data checks. Those extra tools cover the questions that need something outside the company, which saves switching between the assistant and a browser tab mid task.
Pros and cons
Pros
- The community edition is MIT licensed and free to self-host, so you can test the whole thing on one server before talking to sales.
- Connector coverage is broad, with more than 40 apps documented, and adding a custom one is realistic for a team with Python experience.
- Citations on every answer make the output verifiable, which is the single biggest reason teams trust it over a generic chatbot.
- Model choice stays open. You can run GPT, Claude, Gemini, or a local model, and switch when pricing or quality changes.
- The hosted option exists, so a team that doesn't want to run infrastructure can still use the same product.
Cons
- Self-hosting means real operations work. You're responsible for Docker or Kubernetes, upgrades, backups, and the Vespa index, and that's a part-time job for someone.
- Permission sync, SSO, and audit features sit behind the Enterprise plan. On the free tier you get basic roles rather than mirrored document permissions, which rules it out for some regulated teams.
- Large corpora take time and disk to index, and the first sync of a big Drive or Confluence space can run for hours before search feels complete.
- The rename from Danswer to Onyx creates friction. Support threads, blog posts, and download instructions are split across two names, so you'll occasionally land on a stale page.
Frequently asked questions
Yes. The company rebranded the product from Danswer to Onyx in 2024. The code, documentation, and releases now live under the Onyx name on GitHub and at onyx.app, while the old danswer package names and Docker images still show up in older guides.
Related content
Explore related tools, skills, and articles for Danswer.
Danswer Alternatives
Base Code
Base44 (Wix) · Coding · ProductivityBase Code is a shared cloud development environment from Base44. Teammates describe changes to an existing GitHub codebase in chat, preview them live and send them back as pull requests.
Google EmbeddingGemma 2
Google DeepMind · Writing · Image · CodingGoogle EmbeddingGemma 2 is a 740M-parameter open model that turns text, code, images, audio, and video frames into vectors in one 768-dimension space, running on phones and laptops.
GitGlow
GitGlow · Coding · ProductivityGitGlow is a local-first desktop Git client that lets independent developers review their own code and their AI agents' changes in one place before shipping.