AI NewsDev toolsAnnouncement

Papermark adds an MCP server with 57 tools for AI data rooms

Papermark launched Virtual Data Rooms with a public API, a CLI, and an MCP server. The server exposes 57 typed tools to AI agents.

AI News

Editorial2 min read

LinkedInX
The Papermark Virtual Data Room launch image on Product Hunt.

Image: Papermark

Why it mattersA deal workflow that used to need a human can be run by Claude or Cursor. The audit log still records every action.

A data room is a folder of confidential files with a web wrapper around it. Papermark now lets an AI agent build and run one on its own.

Papermark launched its Virtual Data Rooms product on Product Hunt. The company's agents page gives the specifics. The MCP server exposes 57 typed tools. Each tool runs under a scoped access token. Every call is written to an append-only audit log. The protocol is the Model Context Protocol, used by Claude and Cursor.

What the 57 tools do

The MCP server covers eight areas of the product. The tools create data rooms, upload documents, organise folders, and generate secure permissioned links. They set per-viewer or per-group permissions. They read page-by-page analytics on who viewed what. They run the Q&A module for due-diligence questions. They manage webhooks and export the audit trail.

The product itself has been built up over months. The Papermark changelog records a first MCP server with 40 or more tools in late June. AI redaction and a new analytics dashboard arrived in August. Multi-brand support and document thumbnails followed in September. This launch adds the Virtual Data Rooms product itself. The 57-tool MCP server is its agent interface.

Papermark lists Claude and Cursor as the first two agent clients. The tools are typed. An agent sees a schema for each one. It does not have to parse a free-form API response.

What this costs and who it is for

The API, the CLI, and the MCP server are only on Papermark's Business plan. Data Rooms pricing starts at €99 a month on an annual plan. The company says the platform is open source and self-hostable. A team on a different plan can run the server itself.

Co-founder Marc Seitz describes the day-to-day pattern in his Product Hunt post. One example he gives is a create call. "Ask for a buy-side M&A room for Project Atlas." The agent returns the folders, the permissions and the share link. The second example is a read-only query. "Who viewed my pitch deck this week?"

The limits Papermark names itself

Papermark says 80,000 users on Product Hunt. The site names Backtrace Capital, TBD VC, DataPower VC, Two, Differential Bio, and Yuno as customers. These figures and names are the company's own.

Scoped access tokens limit what an agent can touch. The audit log is append-only. A wrong action can be read later but cannot be hidden. A permission set by an agent still goes live before any human reads the log. That is the trade a buyer should think about before letting an agent run a live room.

Source

Papermark Agents and MCP, by Papermark. Launch page: Papermark Virtual Data Room on Product Hunt.

SourcePapermark

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project