Self-host Onyx Community Edition
helm/onyx-0.9.45Open-source AI enterprise search and chat over your apps
irm https://onyx.app/install_onyx.ps1 | iex- Stars
- 33 412
- CPU
- 4 vCPU minimum (Standard mode; Lite 2 vCPU)
- RAM
- 10 GB minimum (Standard mode; Lite 2 GB)
- Storage
- 32 GB + ~2.5x indexed data (Standard; Lite 10 GB)
- Port
- 3000 (HTTP entry point of the Docker Compose deployment (access Onyx at localhost:3000))
- Requires
- PostgreSQL
What is Onyx Community Edition?
Onyx is an open-source AI platform that acts as a knowledge and context layer for a team: it connects to more than 50 applications, ingests documents along with their metadata and permissions, and serves that context to humans and AI agents. Core capabilities are agentic RAG search over a hybrid index, an LLM chat UI, custom agents with actions, deep research reports, live web search, a sandboxed Python code interpreter, image generation and artifacts, plus voice mode. It works with all major LLM providers, self-hosted (Ollama, LiteLLM, vLLM) or proprietary (Anthropic, OpenAI, Gemini), and exposes itself through a web UI, a desktop app, Slack and Discord bots, an MCP server, a Chrome extension and an embeddable website widget. Self-hosting is done with the guided onyx-cli installer or the Docker Compose files under deployment/docker_compose, with Kubernetes, Helm, Terraform and AWS/GCP/Azure/DigitalOcean guides; an air-gapped offline install is supported. Two modes exist: Standard (indexing, workers, model servers, Redis and object store) and Lite, a lightweight chat deployment under 1 GB memory that cannot index documents. The Community Edition is MIT licensed; enterprise extras such as OIDC/SAML SSO, RBAC, white-labelling and data exports sit in separately licensed ee directories and paid plans.
Should you self-host Onyx Community Edition?
Onyx Community Edition scores 79/100 on selfhost.directory (how scoring works) and ranks #5 of 41 in Search Engines. It has 33 412 GitHub stars, +986 in the last 30 days.
It shipped 145 stable releases in the past year. The latest, helm/onyx-0.9.45, came out on 14 November 2026. New versions land roughly every 1 days.
Setup is moderate: expect some configuration beyond a single command. It is MIT-licensed and written mostly in Python.
Its closest alternative on selfhost.directory is pipeshub (66/100, 4 124 stars). Onyx Community Edition scores higher overall. See Onyx Community Edition vs pipeshub side by side.
Features
- AI chat interface
- Air-gapped deployment
- Custom AI agents
- Deep research reports
- Desktop App
- Enterprise search / RAG over connected appsIndexing requires Standard mode; Lite does not index documents.
- Image Generation
- MCP support (server and client actions)
- Query history and usage analytics
- REST API
- Role-based Access ControlRBAC appears in the Business tier of the pricing page.
- Sandboxed code execution
- Single sign-on (OIDC / SAML)Listed under the paid Enterprise plan on the pricing page.
- Voice Mode
- Web SearchMust be configured by an admin.
FAQ
Why is Onyx refusing to search the internet?
Web Search has to be enabled by an admin with a supported search provider; a scraper is optional since one is built in.
Does Onyx keep its knowledge up to date?
Connectors refresh on a schedule, typically every 30 minutes, with documents, metadata and permissions all updated.
Why can't Onyx find information that I know exists?
Indexing may not have finished, access controls may be wrong (the system defaults to least permissive), or a connector is misconfigured.
Responses are really slow, is this expected?
No; slow responses usually come from a slow or non-streaming LLM, or locally configured reranking/embedding models without a GPU.
Why is my Onyx deployment not coming up?
Under-resourcing the deployment is the most common cause; the docs advise checking the Resourcing page requirements.
Can I run Onyx offline?
Yes, the whole system can be air-gapped, but you must run a local LLM, and a recent non-quantized model is recommended.
How do I know if Onyx is right for me?
It targets teams of any size, from enterprises needing unified search to a single user wanting a local chat UI, emphasising privacy and grounded answers.
How is my private data being processed?
The docs point to the Security section, stating Onyx is built with security and data privacy as a core commitment.
What kind of questions can I ask Onyx?
Questions about knowledge from connected applications, general world knowledge the LLM already has, and anything searchable on the web when Web Search is configured.
More in Search Engines
openbeamOpenBeam is a self-hosted, open-source enterprise search platform unifying SaaS tools and IoT systems into one searchable layer with hybrid search and AI agents.
OpenDocumentsOpen-source RAG system you run on your own hardware that indexes files from GitHub, Notion, Drive, Confluence, S3, and local folders, answering questions with citations.
openrecallOpen-source, self-hosted tool that snapshots your screen and makes text and images searchable with local OCR and semantic search — an alternative to Windows Recall.
OpenSearchOpenSearch is a powerful open-source search and analytics suite. Apache 2.0 fork of Elasticsearch 7.10 with security, anomaly detection, alerting, vector DB, and ML.
OpenSERPSelf-hosted open-source SERP API querying six search engines with megasearch, page extraction, and a command-line interface.
panySelf-hosted semantic search engine running in PostgreSQL via pgvector, using CLIP embeddings for text and image search to replace Pinecone or Weaviate.
PgStackPgStack is a self-hosted, open-source tool that creates search and audit read replicas of your Postgres database via logical replication.
Photonan open source geocoder for openstreetmap data
qdrantQdrant is an open-source, self-hosted vector database in Rust for fast AI similarity search with filtering and quantization.
quickwitOpen-source self-hosted search engine for observability: full-text search and aggregations on logs and traces in S3 or MinIO, with an Elasticsearch-compatible API.
sciraSelf-hosted AGPL-3.0 AI search engine that plans multi-step research across 17 modes, pulls live sources, and answers with clickable inline citations.
searchmysite.netsearchmysite.net is a self-hosted, open-source search engine for personal sites. It indexes vetted sites, boosts ad-free content, and offers a privacy-first alternative.

0 comments