Papermark Agents

Product Introduction

1. Definition: Papermark Agents is an AI-native document workflow platform that functions as a Model Context Protocol (MCP) server and REST API. It provides a secure data room environment specifically engineered for autonomous operation by AI agents.
2. Core Value Proposition: The core value is enabling AI agents to execute complete, end-to-end document workflows—from data room creation and file ingestion to secure link distribution and analytics—while maintaining enterprise-grade security, permissions, and a full audit trail. It bridges the gap between AI automation and secure document management.

Main Features

1. MCP Server & REST API (43+ Tools): This is the central interface. It exposes the full functionality of Papermark as a set of typed, machine-readable tools via the @papermark/mcp-server npm package or direct HTTP calls to api.papermark.com. Tools map 1:1 to REST endpoints, allowing agents like Claude, ChatGPT, Cursor, or custom code to perform actions such as create_dataroom(), upload_document(), create_link(), and get_view_analytics() via simple, secure function calls.
2. Scoped Token Security & Permissions: Authentication is handled via pm_live_ scoped bearer tokens created in the user's settings. Every agent action is bound by the permissions of this token. Link-level and folder-level permissions are strictly enforced, ensuring an agent can only access, share, or query documents it is explicitly authorized to view. This includes dynamic watermarking, NDA gating, and email verification on links created by agents.
3. Page-by-Page Analytics & Grounded Q&A: Agents can retrieve granular, structured data on document engagement. Using get_view_analytics and list_views, they can pull per-visitor metrics including time spent on each page, download events, and view history. Furthermore, search_documents enables agents to perform Q&A across a data room, returning answers with direct citations to source pages the token is permitted to access, preventing fabrication.
4. Append-Only Audit Log & Webhook Events: Every action taken by an agent—view, upload, link creation, or access change—is logged immutably in an append-only audit trail, accessible via the get_audit_log tool for compliance and record-keeping. Agents can also subscribe to real-time webhooks (e.g., link.viewed, document.created) to react to events as they happen.

Problems Solved

1. Pain Point: Manual, repetitive, and time-consuming setup and management of secure data rooms for fundraising, M&A, due diligence, and document sharing. This includes tasks like manually creating rooms, uploading and organizing files, generating individual watermarked links, tracking document engagement, and compiling compliance reports.
2. Target Audience: Founders raising venture capital, M&A advisors and investment bankers managing deal processes, private equity and venture capital firms distributing LP updates, real estate brokers managing property listings, SaaS platforms needing to embed secure document sharing, and enterprise sales teams automating proposal tracking.
3. Use Cases: An AI agent can be prompted to "Create a Series B data room, upload the ~/raise/ folder, sort documents into Financials, Legal, Product, and Team folders, and generate a password-protected, email-verified link for each VC on a provided list." Another use is an agent monitoring a deal pipeline that automatically flags investors who viewed a financial deck but haven't responded in five days, drafting personalized follow-ups.

Unique Advantages

1. Differentiation: Unlike traditional virtual data rooms (VDRs) or document sharing platforms (e.g., DocSend, Intralinks, SharePoint) that require manual human operation through a UI, Papermark Agents is built for AI agents first. It provides the API-driven backbone and security guarantees that allow AI to run real workflows autonomously, not just analyze static content. It is also open-source and self-hostable.
2. Key Innovation: The key innovation is the deep integration of the Model Context Protocol (MCP) with a secure document management platform. By providing a comprehensive, typed toolset over MCP, it allows any compatible AI agent to interact with a data room as naturally as it would with a local file system, but with all the security, permissions, and analytics of a managed platform. This creates a new category of "AI agents for secure data rooms."

Frequently Asked Questions (FAQ)

1. How do Papermark Agents handle document security and permissions? Papermark Agents enforce security at every step. Agents authenticate with scoped pm_live_ bearer tokens, and all actions respect link-level and folder-level permissions defined in the data room. Documents are protected with dynamic watermarking, screenshot protection, and options for NDA gating and email verification. Every action is logged in an immutable audit trail.
2. What AI models and clients can I use with Papermark Agents? You can use Papermark Agents with any MCP-compatible AI client, including Claude Desktop, Claude Code, Cursor, ChatGPT (via Connectors), and custom-built agents. The @papermark/mcp-server works via stdio for local clients, or you can use the remote HTTP endpoint at mcp.papermark.com for browser-based clients. The same 43 tools are available across all interfaces.
3. Can an AI agent answer questions about the contents of my data room? Yes. The search_documents tool allows an agent to perform grounded question-answering across all documents in a data room it is authorized to access. The agent returns answers with direct citations back to the specific pages in the source documents, ensuring accuracy and verifiability.
4. How does this compare to using a standard API for document sharing? While a standard API provides programmatic access, Papermark Agents is optimized for AI interaction through the MCP standard, offering typed tools that are easier for agents to discover and invoke. It packages a complete workflow solution—not just file storage—with built-in analytics, granular security, and audit trails specifically designed for high-stakes document processes like fundraising and M&A.