onyx-dot-app/onyx-foss

Open Source AI Platform - AI Chat with advanced features that works with every LLM

Onyx – Open‑source “context layer” for teams and AI agents

Onyx is a self‑hosted platform that gathers data from dozens of workplace apps (Slack, Google Docs, GitHub, etc.), builds a searchable knowledge store, and makes that knowledge available to large language models (LLMs) and AI agents. It is positioned as the “knowledge/context layer” that lets an LLM answer questions with up‑to‑date, private company information instead of only public data.

What the repo provides

  • Core server code (MIT‑licensed) that runs the indexing pipeline, vector/keyword stores, and the API that agents and UI components talk to.
  • Connectors for >50 SaaS applications that pull documents, metadata and permissions into the internal index.
  • Agentic RAG stack – a hybrid vector‑plus‑keyword index plus a custom retrieval‑agent that selects the most relevant chunks for a query.
  • Web UI & desktop client for chatting with the system, creating custom agents, and viewing generated artifacts.
  • Integrations – Slack/Discord bots, a Chrome extension, and an embeddable widget so you can query Onyx from anywhere.
  • Deployment scripts for Docker, Kubernetes, Helm/Terraform and cloud‑provider guides.
  • Two deployment flavours – a full‑featured “Standard” mode (vector store, background workers, model inference) and a lightweight “Lite” mode (just the chat UI, <1 GB RAM, no indexing).

Key capabilities

Feature What it does
Agentic RAG Retrieves the most relevant internal documents and feeds them to an LLM, reducing token usage and improving answer accuracy.
Deep Research Multi‑step workflows that let agents gather, synthesize, and cite information across many sources.
Custom Agents Define agents with a specific knowledge subset, custom prompts, and the ability to invoke actions.
Web Search Optional live web lookup via Serper, Google PSE, Brave, SearXNG, Firecrawl, Exa, etc.
External Actions / MCP Agents can call APIs of connected apps (e.g., create a Jira ticket, update a Notion page) to complete tasks end‑to‑end.
Secure Sandbox Isolated execution environment for running code or processing intermediate artifacts safely.
Artifacts Generate PDFs, images, or other downloadable files as part of a workflow.
Voice Mode Speech‑to‑text and text‑to‑speech interface for hands‑free interaction.
Fine‑grained permissions Role‑based access control, SSO (Google, OIDC, SAML), SCIM provisioning, and audit logs.

Typical use cases

  • Internal Q&A – Employees ask the chat bot about company policies, product specs, or recent project updates and receive answers backed by the latest internal docs.
  • Automated workflows – An agent reads a support ticket, looks up relevant troubleshooting steps, and creates a resolution ticket in the ticketing system.
  • Research reports – The system gathers data from multiple sources, synthesizes a report, and outputs a PDF or slide deck.
  • Developer assistance – Connectors pull code snippets, design docs, and issue trackers so an LLM can answer technical questions with the correct context.

Getting started

# One‑liner install (Docker‑based) provided by the project
curl -fsSL https://onyx.app/install_onyx.sh | bash

After installation you can:

  1. Run the Docker compose stack (or Helm chart for Kubernetes).
  2. Configure connectors to the apps you want to index.
  3. Add your LLM provider (OpenAI, Anthropic, Ollama, etc.) and an embedding model.
  4. Open the web UI (default http://localhost:3000) and start chatting.

Licensing & editions

  • Community Edition (CE) – Fully open source under the MIT license; includes core RAG, chat UI, agents, and most connectors.
  • Enterprise Edition (EE) – Proprietary add‑ons for large organisations (advanced analytics, white‑labeling, extra security controls). The repo only contains the CE; EE features are documented on the commercial site.

Who should look at this repo?

  • Teams that need private, self‑hosted AI assistance and cannot rely on public LLM knowledge bases.
  • Developers who want to build custom AI agents that act on internal tools.
  • Organizations that require data sovereignty, air‑gapped deployments, or strict RBAC.

Bottom line: Onyx is a genuine open‑source platform that stitches together workplace data, a vector/keyword index, and LLMs to deliver fast, context‑aware AI assistance. It ships with connectors, UI, and deployment tooling, making it a practical starting point for building internal AI copilots.

Related

  • Project
    onyx-dot-app/onyx

    Onyx is an open‑source, self‑hosted platform that indexes data from 50+ workplace apps, builds a hybrid vector‑keyword store, and exposes that knowledge to LLMs via agents. It supports web search, sandboxed code execution, external actions, voice interaction, and works with any major LLM provider. Deployable via Docker/Kubernetes (standard or lightweight Lite mode) and offered under an MIT‑licensed Community Edition, with an Enterprise add‑on for larger teams.

  • Project
    ogx-ai/ogx

    OGX is an open‑source, model‑agnostic API server that mimics OpenAI’s REST endpoints. It lets you run any LLM (Llama, GPT, Gemini, etc.) locally or in the cloud, while keeping the same client code. In addition to standard chat/completion/embedding calls, OGX offers a “Responses” API for tool‑calling and RAG, versioned skill bundles, and native routes for Anthropic and Google GenAI SDKs. Install with a one‑line script, start the server, and point any OpenAI‑compatible client at it.

  • Project
    brainlid/langchain

    A framework for Elixir applications to integrate LLMs and chain them with external data and custom functions to build agentic, data-aware AI applications.

  • Project
    the-open-agent/openagent

    An open-source personal AI assistant platform that combines LLMs, RAG, and autonomous agent loops into a single self-hostable binary.

  • Project
    java-up-up/nexus-agent

    An enterprise-grade AI agent platform that implements a complete loop from document governance and hybrid RAG retrieval to ReAct-based agentic reasoning and MCP tool integration.