AI for your data room
Let investors and buyers ask your documents questions, build rooms in seconds, and run Papermark from Claude, ChatGPT or Perplexity. With the permissions and audit trail of the secure, modern data room.
04 Legal
Ask this data room
Six ways AI works in Papermark
For the people you share with, and for the team running the room
Chat with a data room
Visitors ask questions across every document they can see and get answers that cite the file, folder and page.
See how it works →Chat with a document
Summarise a deck, pull figures out of a spreadsheet or turn a pitch deck into a memo, with page references.
See how it works →Generate a data room
Describe the deal and AI drafts the folder structure for M&A, fundraising, real estate or legal work.
See how it works →AI redaction
Find names, emails, account numbers and tax IDs on every page, review each hit, and burn the approved ones in.
See how it works →Agents: MCP, API and CLI
60+ tools that let AI agents create rooms, upload files, share links and read analytics, with scoped tokens.
See how it works →Claude, ChatGPT and Perplexity
Connect Papermark to the assistant you already use and run your data room from a prompt.
See how it works →Give visitors an AI that knows the room
And decide exactly who gets it, in three switches
A buyer asks which contracts have a change of control clause, and the answer comes back with the file, folder and page. An investor asks for a summary of the financials instead of opening twelve documents. The chat only ever answers from documents that visitor can already open.
01
Team
Settings → AI
An admin turns on AI for the team. Nothing is indexed until you do.
02
Data room or document
AI Agents card
Enable AI for one room or file and index it. Re-index after changes, or disable it and the index is deleted.
03
Link
Link settings
Switch AI on for the links that should have it, so one buyer group can chat while another can't.
Describe the deal
AI drafts the data room
Start from M&A due diligence, fundraising, real estate, legal or your own description, in any language. AI proposes up to eight top-level folders with subfolders. Rename, reorder or regenerate before you create the room, then drag your files in.
Top-level folders
Up to 8
Sent to the AI
Only your prompt
Describe your data room
Sell-side M&A for a B2B SaaS company with customers in the US and Europe
01 Corporate and governance
02 Financials
03 Commercial
04 Legal
05 People
06 Technology and IP
Run Papermark from Claude, ChatGPT and Perplexity
One MCP server, 60+ tools, every agent
Papermark Agents connect your workspace to the assistant you already use. Ask it to build a room, upload the latest deck, share a link per fund or report who read what, and it works on your real documents, data rooms and analytics.
Claude
“Set up a data room for our Series A with folders for the deck, financials, cap table and legal docs, and share it with email verification.”
Create a data room with Claude →Perplexity Computer
“Every Monday, check my data room analytics and draft follow-ups to investors who viewed the deck twice but haven't replied.”
Papermark on Perplexity Computer →ChatGPT and Cursor
“Which investors opened the deck this week, where did they spend the most time, and who stopped at the financials?”
Connect any MCP client →Hosted connector
For claude.ai, ChatGPT, Cursor and any MCP client. Sign in with OAuth or a token.
https://mcp.papermark.com/mcp
Local, for Claude Desktop and Claude Code
Run the open package with an API token from Settings → API Keys.
npx -y @papermark/mcp-server
AI with the controls of a data room
Nothing reaches a model or an agent that you did not allow
Permissions still apply
The chat only answers from documents the visitor is allowed to see. Folder and group permissions carry over.
No training on your data
Chat runs on OpenAI's API, which does not use data sent through it to train models.
Off means deleted
Disable AI for a document and its index is deleted. Only the description, never your files, goes into room generation.
Scoped, revocable tokens
Give an agent full access, read-only access, or read and write per resource. Revoke a token at any time.
Every agent call logged
Actions taken over the MCP server land in the audit log next to every view and download.
Nothing applied without review
AI redaction proposes; you accept or decline each hit before a new version is created.
Read how we protect your documents on the security page.
