What is Onyx?
Onyx is an open-source AI platform that connects organizational knowledge to AI chat, search, and agents. Onyx serves individual users, teams, and enterprises that need answers grounded in information from their own applications. The project describes Onyx as a knowledge and context layer for both people and AI agents.
Onyx connects to 50+ applications and builds a searchable representation of their knowledge. The ingestion process includes data, metadata, and permissions, allowing downstream queries to retrieve relevant documents with access controls. Beyond internal search, Onyx supports research workflows, web information, and agents that can perform external actions.
What Onyx can do
Onyx combines knowledge retrieval with tools for research and task execution. Onyx uses a hybrid index and an agent harness designed for information retrieval, while additional capabilities let users extend internal knowledge with web results, execute code in a sandbox, and create downloadable outputs.
- Onyx retrieves knowledge through agentic RAG using a hybrid index.
- Onyx creates in-depth reports through multi-step research.
- Onyx builds custom agents with selected knowledge and instructions.
- Onyx searches the web through supported search providers.
- Onyx lets agents take actions in external applications through external actions and MCP.
- Onyx executes code in a sandbox and generates downloadable artifacts.
- Onyx supports voice interaction through speech-to-text and text-to-speech.
How Onyx works
Onyx ingests information from connected sources and filters retrieved context to relevant documents for each query. Onyx supports model providers including Ollama, LiteLLM, vLLM, Anthropic, OpenAI, and Gemini. Organizations can choose their embedding model and LLM provider, including models that run locally within a self-hosted deployment.
Onyx supports Docker, Kubernetes, Helm, and Terraform deployments, with guides for major cloud providers. Onyx offers web and desktop access, Slack and Discord bots, a Chrome extension, and an embeddable widget. An MCP server also supplies company context to Claude Code, Open Code, Codex, and other MCP clients under the user's access controls.
Who Onyx is for
Onyx serves users ranging from individuals to large enterprises. Teams can use shared chats and agents to work with organizational knowledge, while developers can connect MCP-compatible assistants to company context. Organizations handling sensitive information can use self-hosted, air-gappable deployments in which the document index, database, and processing remain within a self-contained set of services.
- Onyx helps teams find answers across connected organizational applications.
- Onyx supports research tasks that combine internal knowledge with live web search.
- Onyx gives coding assistants access to company context through MCP.
- Onyx answers organizational questions directly in Slack or Discord.
Onyx pricing and editions
Onyx Community Edition is freely available under the MIT license and includes core RAG, AI chat, agents, and actions. Onyx Enterprise Edition includes additional features primarily intended for larger organizations. Onyx Cloud provides a way to try Onyx for free without deploying the software, while the repository provides a single-command self-hosted installation path.
Onyx strengths and deployment limitations
Onyx combines broad application connectivity, multiple model-provider options, and deployment control. Onyx Standard includes the complete feature set, including document indexing and background synchronization components. Onyx Lite reduces resource requirements and deployment complexity for users focused on the chat interface and agents, but that lighter configuration does not provide document indexing.
- Onyx Lite cannot index documents.
- Onyx Standard requires a more complex stack than Onyx Lite.
- Onyx separates additional enterprise-oriented capabilities into Enterprise Edition.
Onyx FAQ
What is Onyx?
Onyx is an open-source AI platform and knowledge layer for teams and AI agents. Onyx indexes information from connected applications and makes organizational knowledge available through AI chat, enterprise search, research workflows, and custom agents.
Is Onyx free?
Onyx Community Edition is freely available under the MIT license and includes core RAG, AI chat, agents, and actions. Onyx Cloud also offers a way to try Onyx for free without deploying the software.
Can Onyx be self-hosted?
Onyx supports self-hosted deployments using Docker, Kubernetes, Helm, and Terraform. Onyx can operate in an air-gapped environment, with its document index, database, and processing contained within its services, and supports locally running embedding models and LLMs.
What is the difference between Onyx Standard and Onyx Lite?
Onyx Standard provides the complete feature set, including document indexing, connector synchronization workers, model inference servers, caching, and blob storage. Onyx Lite uses a less complex stack with under 1GB memory requirements but cannot index documents.
Does Onyx support MCP and integrations?
Onyx provides an MCP server for Claude Code, Open Code, Codex, and other MCP clients, supplying company context under the user's access controls. Onyx also offers Slack and Discord bots, a Chrome extension, an embeddable widget, and external actions.
Which language models does Onyx support?
Onyx supports self-hosted model providers such as Ollama, LiteLLM, and vLLM, as well as proprietary providers including Anthropic, OpenAI, and Gemini. Onyx also allows organizations to choose their embedding model and LLM provider.
From the community
No reviews yet
Be the first to review this tool. What you write helps the next person choose.




