About Onyx 4.7.1 — Open-Source AI Platform for Team Knowledge
Onyx, formerly Danswer, is a self-hostable AI platform for teams. It gives you a chat interface that works with any major model, connected to your organisation’s knowledge through more than 50 connectors, with RAG, deep research, agents, web search and code execution built in. The Community Edition is MIT licensed.
This page links to files from Onyx’s GitHub releases: the v4.7.1 source archive, the Onyx desktop app from that release, and the Onyx command-line client. Bineret did not build Onyx and is not affiliated with it.
What it does
- Agentic RAG. Search and answers over a hybrid index, with agents doing the retrieval.
- Deep Research. Multi-step research that produces an in-depth report.
- Custom agents with their own instructions, knowledge and actions.
- Web search through Serper, Google PSE, Brave, SearXNG and others, with its own crawler and support for Firecrawl and Exa.
- Artifacts. Documents, graphics and other files you can download.
- Actions and MCP, so agents can work with outside applications.
- Code execution in a sandbox, for data analysis, graphs and file edits.
- Voice mode and image generation.
- Any model. Self-hosted, such as Ollama, LiteLLM and vLLM, or hosted, such as Anthropic, OpenAI and Gemini.
Lite or Standard
Onyx deploys in two modes.
Onyx Lite is a lightweight chat UI with agents. It needs under 1 GB of memory and suits a quick trial, or a team that only wants chat and agents.
Standard Onyx is the full feature set. On top of Lite it runs a vector and keyword index for RAG, background workers that sync knowledge from connectors, model inference servers used during indexing, Redis as a cache and MinIO as a blob store. It deploys with Docker, Kubernetes, Helm or Terraform, with guides for the major clouds.
The project’s one-command installer, curl -fsSL https://onyx.app/install_onyx.sh | bash, downloads the Onyx CLI and runs its guided deployment, with options such as --lite and --dry-run. It is a short script; read it before you pipe it into a shell.
Community and Enterprise
The Community Edition is MIT and covers the core: chat, RAG, agents and actions. Code in the ee directories (backend/ee, web/src/app/ee and web/src/ee) is under the separate Onyx Enterprise License. The README’s enterprise section lists sharing, single sign-on through Google OAuth, OIDC or SAML with SCIM provisioning, role-based access control, analytics, query history, custom code hooks and whitelabelling. Check Onyx’s pricing page for which of those need the Enterprise Edition before you plan around them.
What the files are
- The source archive of v4.7.1, 27.4 MB and 7,564 files, released on 8 September 2026. Deployment files and guides live in it.
- Onyx Desktop, a lightweight Tauri app of 4 to 12 MB. Its README describes it as a desktop app for Onyx Cloud: a native window with keyboard shortcuts, several windows at once, and a remembered window size. It is a client, not an Onyx server.
- onyx-cli 1.4.2, a terminal client for querying knowledge in Onyx. It has an interactive chat interface for people and non-interactive commands for scripts and AI agents, authenticates with a personal access token, and also runs the guided deployment the installer uses.
When to use something else
If you are one person chatting with your own documents on a desktop, AnythingLLM is much less to run. If your documents are scans and complex layouts and parsing quality matters most, look at RAGFlow. Onyx fits best where a team wants one AI interface over many company systems.
Key capabilities
Answers from your own systems
More than 50 connectors index what your team already uses, and agentic RAG over a hybrid index answers from it. Deep Research runs a multi-step flow when a question needs a full report.
Lite or Standard
Lite is chat and agents in under 1 GB of memory. Standard adds the RAG index, connector sync workers, model servers, Redis and MinIO, and deploys on Docker, Kubernetes, Helm or Terraform.
Two editions, one repository
The Community Edition is MIT. Code in the ee directories is under the Onyx Enterprise License, and organisation features such as SSO, SCIM and RBAC are listed on the enterprise side; check which need a paid plan.
Desktop app and CLI
The release includes a 4 to 12 MB Tauri desktop client for Onyx Cloud, and onyx-cli, a terminal client with a chat interface and scriptable commands that also drives the guided deployment.
Any model, including local
Self-hosted providers such as Ollama, LiteLLM and vLLM, or hosted ones such as Anthropic, OpenAI and Gemini, with web search, code execution, artifacts, voice and image generation on top.
Questions & Answers
Is Onyx free?
The Community Edition is MIT licensed and covers chat, RAG, agents and actions. Code under the ee directories is under the Onyx Enterprise License, and some organisation-level features need the paid Enterprise Edition. Onyx's pricing page lists which.
Is Onyx the same project as Danswer?
Yes. The licence's copyright line still names DanswerAI, Inc.
Lite or Standard?
Lite runs in under 1 GB of memory and gives you chat and agents. Standard adds the RAG index, connector sync workers, model servers, Redis and MinIO. If you want answers from your company's documents, you want Standard.
Does the desktop app include a server?
No. It is a lightweight client, described in its README as a desktop app for Onyx Cloud. It does not index documents or run models itself.
Can it use local models?
Yes. Onyx supports self-hosted providers such as Ollama, LiteLLM and vLLM alongside hosted ones.
Is the one-line installer safe to run?
It downloads the Onyx CLI from GitHub and starts a guided deployment. Download it and read it first, then run it with --dry-run to see what it would do before letting it change anything.
Is Bineret connected to Onyx?
No. This page describes Onyx and links to files on the project's GitHub releases.
Tutorials
Try the guided installer
Onyx's own route installs the Onyx CLI and runs a guided Docker deployment. Read the script first, then do a dry run:
curl -fsSL https://onyx.app/install_onyx.sh -o install_onyx.sh
less install_onyx.sh
bash install_onyx.sh --dry-run
bash install_onyx.sh # or: bash install_onyx.sh --lite--lite deploys Onyx Lite, without OpenSearch, Redis or model servers. --dir sets where the deployment lives; the default is ~/.config/onyx. --tag picks the image version, and defaults to the latest release.
Deploy from the source archive
Unzip onyx-4.7.1.zip. The deployment folder holds the Docker, Kubernetes and Helm material; follow Onyx's deployment guides for the mode you want, since Standard and Lite differ in what they start.
Connect a model and your knowledge
In the admin panel, add a model provider: a self-hosted one such as Ollama, LiteLLM or vLLM, or a hosted one such as Anthropic, OpenAI or Gemini. Then add connectors for the systems your team uses, so their content is indexed for search and answers.
Install the desktop app
Windows: run Onyx_x64.exe. macOS: open Onyx_universal.dmg, which runs on both Apple Silicon and Intel. Debian or Ubuntu:
sudo apt install ./Onyx_amd64.debThe app is a client; it needs an Onyx server, such as Onyx Cloud, to talk to.
Use the CLI
Unpack the archive for your system, or install it with pip:
pip install onyx-cli
onyx-cli chatOn first launch it asks for your Onyx server URL and a personal access token, tests the connection, and saves them to ~/.config/onyx-cli/config.json. For scripts, set ONYX_SERVER_URL and ONYX_PAT instead. Without a server URL it defaults to https://cloud.onyx.app.
Support
Where support comes from
Onyx is developed by its company and contributors. Start with the documentation, report bugs as GitHub issues, follow the security policy for vulnerabilities, and use the project's Discord community for questions.
What Bineret covers
This page. If a download link is broken or something written here is wrong, tell us and we will fix it. We do not deploy, host or configure Onyx.
Enterprise Edition and Onyx Cloud
Both are sold by Onyx, not by Bineret. Licensing and pricing questions go to Onyx directly through its website.
No warranty
The MIT licence provides the Community Edition as-is. Connectors give Onyx read access to real company systems, so grant them the narrowest permissions that work.
