companyRAG
With the companyRAG addon, you can make large volumes of documents available to your agents via the MCP interface. Documents can be indexed from different sources and integrated using the RAG - Retrieval Augmented Generation methodology. There is no limit on the length of individual documents or their total number.
Indexing can be implemented as a one-time or recurring process, depending on your use case. In addition to classic semantic search, companyRAG can also extract structured data from the same documents and make tabular datasets queryable via SQL.
companyRAG is available to appropriately authorized users in the Knowledge area.
Types of collections
Section titled “Types of collections”All content in companyRAG lives in collections. The type of a collection determines how its content is stored and queried. It is shown as a badge on every collection in the overview.
| Type | Content | Query |
|---|---|---|
| Documents | Documents from upload, SharePoint, Nextcloud, Google Drive, GitHub or web crawl | Semantic search and full-text search; optionally also structured fields |
| Dataset | Tables from CSV, Excel and Parquet files, or via an ingestion endpoint | SQL queries across one or more tables, including joins |
| OKF bundle | Markdown knowledge collections in the Open Knowledge Format in GitHub or SharePoint | Same as a document collection |
companyRAG user interface
Section titled “companyRAG user interface”The user interface allows you to add single or multiple files, or entire data sources, for indexing. The interface is divided into the following areas:
- Collections – storage locations for documents and permissions
- Files – all indexed files with their status
- Upload – index single or multiple files manually
- Sources – automatic synchronization from connected systems
- Integrations – connections to SharePoint, Nextcloud, Google Drive, GitHub and web crawl
- Jobs – indexing jobs and their status
- API keys (administrators only)
- Audit log (administrators only)
Further topics
Section titled “Further topics”- Structured data from documents – extract typed fields from recurring document types
- Dataset collections – import tables or fill them live via an endpoint
- OKF bundles – knowledge collections as Markdown in GitHub or SharePoint
- Using companyRAG in CompanyGPT – the
ai-searchMCP server, its tools and agents