Outmind connects to your file servers, Windows shares, and NAS devices, indexes the text in your documents, and makes it searchable via the search function and an AI assistant that cites its sources. The files remain in place—they stay on the server, with their access permissions intact. The connector is ideal for document-intensive businesses whose project history is stored on network drives rather than in the cloud.
What Is Indexed
Sources: network drives (company file servers), local drives. The connector is available in the Outmind app for Windows and Mac; it cannot be configured via the web interface.
Formats whose content and metadata are indexed:
- Text documents: doc, docx, dot, dotx, docm, odt, rtf, txt, md, html
- Spreadsheets: xls, xlsx, xlsb, xlsm, xltx, csv, ods
- Presentations: ppt, pptx, pps, ppsx, pptm, odp
- PDFs, including scanned PDFs (OCR automatically applied to pages without text)
- Email archives: msg, pst, eml
- Compressed archives: zip, rar (contents extracted and indexed file by file)
Formats for which only the metadata is indexed (name, path, dates, size, author, permissions): all other files, including 3D plans and models (dwg, dxf, ifc, rvt, skp), which can be found by name and filtered by type.
Excluded: hidden files, Office temporary files (~$), unsynced cloud files.
An email archived in a .pst file—which is stored on an old server—is opened, and its attachments are read and indexed without having to manually unarchive it.
How It Works
- Scope. The administrator creates the source from the Outmind application, installed on a workstation or a VM, using a service account that has read access to the specified root folders. The scope can then be narrowed or expanded by modifying these root folders, without having to recreate the source.
- Indexing. Outmind reads the text of documents and their metadata, including access permissions (DACL). Only the extracted text content is stored in the index, which is encrypted end-to-end and at rest. The first documents appear within a few minutes; the first full synchronization takes several hours, depending on the volume.
- Incremental synchronization. After the initial run, only new files, modified files, and deleted files are processed (“delta sync”), every 24 hours by default; this frequency can be adjusted. A file deleted from the server is removed from the index during the next run.
- Access Rights. Each user links their own account to the shared resource. In Outmind, users see only the documents to which they already have access on the server; permissions are never modified. In the event of a loss of access to the service account (EPERM error), the “Protect Against Loss of Access” option prevents unintentional mass deindexing; it is disabled when current permissions must be strictly reflected.
- What remains on-site: the files, the directory structure, the access control lists (ACLs), and the server itself. Outmind does not replace either the server or the document management system; it is built on top of them.
What you get
- Full-text and semantic search in the title, body, and metadata, with filters by format, date, source, and operators (for example, “employment contract NOT fixed-term contract”). The Pages tab lists every page where the word appears, allowing you to open the correct section directly.
- Preview documents without opening them, while respecting access rights.
- AI assistant with sources: a question in natural language, an answer generated from accessible documents, with each quoted passage linked to its original file. Works on OCR-processed scans and on files extracted from archives.
- Extraction and processing: summarize, extract the companies mentioned, identify dates or clauses, and anonymize a document.
- A single search across the file server, SharePoint, Teams, Outlook, Google Drive, iManage, Notion, and Slack—each with its own access rules.
- Knowledge made available to other AI systems within the company via MCP, coming soon.
Frequently Asked Questions
Can you connect a Windows file server or a NAS to an AI without moving the documents?
Yes. The Outmind connector reads network shares from the application installed on a workstation or VM, indexes the text and metadata, and leaves the files on the server. The AI assistant responds based on this index and references the original files.
Are the server's access rights being respected?
Yes. Outmind retrieves the permissions (DACL) for each file, and each user links their own account to the source. They can only see what they already have access to on the server; Outmind does not modify any permissions.
Are scanned PDFs and archives included?
Yes. OCR is automatically applied to PDF pages that do not contain text. ZIP, RAR, PST, EML, and MSG files are opened and their contents are indexed, including any attachments.
What happens to a file that has been deleted or moved on the server?
Incremental synchronization removes deleted files from the index during the next run. A folder removed from the source's root folders is gradually deindexed.
What are the infrastructure requirements?
A Windows or Mac workstation or VM with the Outmind application, a service account with read permissions for the folders to be indexed, and network access to the shares. Any changes to the source must be made from the same machine and the same account.
Can we start on a limited scale?
Yes. The source is created based on a few representative root folders; the scope is then expanded by adding folders, without recreating the source or reindexing the existing data.
Test it on your files
Choose two or three representative folders from your server, including archives and scans. We'll index them, and you can ask your own business-related questions about the results. Request a demo.
