Email.
Microsoft 365 and Exchange, Google Workspace, shared mailboxes, with attachments.

it just works.How it works.
Start in RPlay chat. Ask for a document, a summary or a draft. RPlay searches the sources assigned to you and brings the relevant information together.
An example.
What payment terms did we agree with Aerts Logistics, and does the last invoice match?
The signed offer of 12 January 2026 says 30 days end of month. The email thread of 9 February moved that to 45 days for 2026. Invoice 2026-0088 was booked at 30 days: it does not match the agreed 45.
Find the agreed payment terms for this client and check the related invoice.
RPlay can look across the email conversation, the contract in a shared folder and the accounting record. The available sources are agreed during setup.
How it fits together.
You ask in RPlay chat or in an AI app you already use. RPlay Vault, on the Mac Studio in your office, finds the answer across your information sources and sends back only what you may see.
The app you ask in
RPlay chat · Claude Desktop · ChatGPT
1You ask a normal question.
RPlay Vault · your Mac Studio
2Only permitted results go back.
Your information sources
3Cloud or on-premise, as they are today.
Your information sources.
Microsoft 365 and Exchange, Google Workspace, shared mailboxes, with attachments.
SharePoint, OneDrive, file servers and NAS: Office documents, PDFs and scans.
Odoo, Business Central, Salesforce, and sector-specific case or practice software.
Tasks, meeting notes, site reports and FAQ, kept on your RPlay box.
Examples to discuss. We confirm your software, versions and access before the demo.
Apps connect to RPlay through MCP (Model Context Protocol), an open standard; RPlay is the MCP server for the company and enforces each person’s access. Claude Desktop, ChatGPT and Gemini are examples to discuss; support is checked per app. Structured forms use the separate MCP Apps capability.
Connectors: Microsoft 365 through Graph, IMAP mailboxes and archives with attachments, SMB shares, NAS, iCloud Drive and old .msg archives; business systems through the APIs they provide, with read permissions. An internal API or web service of your own can get a connector too. Optional public company research is on request and metered; customer content is never used for it.
Read the official MCP specification.Behind the scenes.
Standard answers come from the AI on your Mac Studio. For a harder question you switch on Deep thinking: RPlay chat sends it to a heavier model through OpenRouter, such as GLM, with only the request and the passages it needs. You see which mode answered. A more powerful local AI server is the alternative that keeps even that in the office.
Start with a question from your working day.