OpenRAG
open-rag.ai
- Free plan
- Web
- API
- Self-hosted
What OpenRAG is
OpenRAG is a modular framework for building document-grounded retrieval-augmented generation systems. It combines semantic search with BM25 keyword matching and multilingual reranking; answers cite the source document and page, with links to the cited location. Supported files include text, office documents, email, audio, video, and image formats, with PDF layout awareness, OCR, image captioning, and audio transcription. The product includes an admin console, chat interface, and OpenAI-compatible API, and can connect to user-provided models or a hosted provider. Partition isolation, owner, editor, and viewer roles, and token or OpenID Connect authentication are among its access controls. Self-hosted under AGPL-3.0, it runs on infrastructure controlled by the deployer, keeping documents, embeddings, and queries within that environment. The self-hosted edition is free with no features held back. Listed prerequisites are Docker, Docker Compose, and at least 16 GB of RAM; a CPU-only profile is optional. CSV, ODT, and HTML support, tool calling, agentic RAG, MCP, and encryption in transit and at rest are listed as coming soon.
Who it suits
OpenRAG suits teams building document-based assistants, legal search, or multimodal enterprise question answering. It is a fit for deployers prepared to run Docker and Docker Compose and manage the underlying infrastructure.
Good to know, and what to watch
- Answers link to cited document pages
- Hybrid search includes multilingual reranking
- Partition roles provide knowledge-base access controls
- Self-hosted edition has no features held back
- Deployment requires Docker and Docker Compose
- Minimum listed memory is 16 GB
- Several formats and features are coming soon
Plans
Every plan and priceManaged OpenRAG service
OpenRAG (AGPL-3.0 self-hosted)
Free plan
The verdict
OpenRAG offers a free, self-hosted framework with document citations and a broad set of parsing and access-control features. Deployers should account for its stated infrastructure requirements and features still listed as coming soon.
What the maker publishes
- Access control
- Partitions isolate knowledge bases; partition roles include owner, editor and viewer, and API tokens are stored as SHA-256 hashes. source
- Access controls
- It supports partition isolation, owner/editor/viewer roles, and token or single sign-on authentication through OpenID Connect providers. source
- Answers
- Answers cite the source document and page, with links that open the source at that page. source
- Answers
- Answers cite the source document and page, with a link that opens the cited page. source
- Audience
- The site describes use for AI assistants, legal search and multimodal enterprise question answering, and says public administrations and private companies use it. source
- Commercial offering
- The maker says there is no paid edition and no feature held back; it also links to a LINAGORA-managed service. source
- Compliance and hosting
- OpenRAG can run on customer infrastructure or hosting qualified SecNumCloud; the site says traceable answers and auditable code provide evidence relevant to GDPR and the AI Act, with obligations remaining with the deployer. source
- Deployment and data
- OpenRAG runs on infrastructure controlled by the deployer, keeping documents, embeddings, and queries within that perimeter. source
- Deployment requirements
- The maker lists Docker and Docker Compose as prerequisites, with 16 GB RAM minimum and an optional CPU-only profile. source
- Document handling
- It supports multimodal parsing including PDF layout awareness, OCR, image captioning, and audio transcription. source
- Document processing
- It supports multimodal parsing with audio transcription, image captioning, OCR and PDF layout awareness. source
- Founded
- 2000 source
- Integrations
- The site names Open WebUI, LangChain, n8n, and Twake.ai as integrations. source
- Integrations
- The product documents compatibility with OpenAI API clients and names Open WebUI, LangChain, n8n and Twake.ai as integrations. source
- Interfaces
- The product includes an admin console, a chat interface, and an OpenAI-compatible API. source
- License
- OpenRAG is licensed under AGPL-3.0. source
- Model choice
- The site says users can connect their own models, including Mistral, Qwen, Lucie, Claude and GPT, or use a hosted provider. source
- Notable limit
- CSV, ODT and HTML support, format-specific chunkers, tool calling, agentic RAG, MCP, and encryption in transit and at rest are listed as coming soon. source
- Notable limits
- The maker lists CSV, ODT, and HTML support, tool calling, agentic RAG, MCP, and encryption in transit and at rest as coming soon. source
- Product
- OpenRAG is a modular framework for building document-grounded retrieval-augmented generation systems. source
- Purpose
- OpenRAG is a modular framework for building document-grounded retrieval-augmented generation systems. source
- Scaling
- Ray distributes ingestion, chunking and embedding across worker nodes for horizontal scaling. source
- Search
- Retrieval combines semantic search with BM25 keyword matching and multilingual reranking. source
- Search and scale
- It supports hybrid search, reranking, and distributed processing with Ray worker nodes. source
Checked against the maker's pages on 3 Oct 2026.