04 Jun 2026

9 min

MCP

Docusign IAM

Connecting AI Agents to Docusign: The MCP Integration Guide

AI agents are changing how teams interact with business systems. Instead of navigating interfaces, clicking through menus, and building search queries, users ask questions in natural language: "What agreements do we have expiring this quarter?" or "What's the status of the MSA we sent to DataTech?" The agent handles the system interaction.

Model Context Protocol (MCP) is the open standard that makes this work. Developed by Anthropic, MCP defines how AI agents discover and use tools from external services. Docusign has released a Beta MCP server that exposes agreement data and workflow capabilities to AI systems - connecting the Docusign IAM platform to Claude, Copilot, and other MCP-compatible agents.

This guide covers what MCP is, how the Docusign MCP server works, how to set it up, what it can and can't do today, and how to think about security.

What Is MCP and Why It Matters

Model Context Protocol is an open standard that defines how AI agents connect to external tools and data sources. Think of it as a universal adapter between AI systems and the applications they need to interact with.

Without MCP, every AI integration is custom. Connecting Claude to Docusign requires custom code. Connecting it to Salesforce requires different code. Connecting to 10 systems means 10 custom integrations, each breaking when APIs change.

With MCP, you build one server for each service. Any MCP-compatible AI agent can discover and use the tools that server exposes. One standard, unlimited connections.

An MCP server exposes three types of capabilities:

Tools - actions the AI agent can perform (search agreements, check status, trigger workflows)
Resources - data the AI agent can read (agreement content, template lists, user information)
Prompts - pre-built templates for common interactions, standardizing how certain requests are handled

The Docusign MCP Server (Beta)

Docusign's official MCP server is available in Beta. It connects to the core IAM APIs and exposes agreement capabilities to AI agents.

What It Can Do Today

The Docusign MCP server bridges three APIs:

Agreement Manager API - agreement repository access:

Search agreements using natural language filters
Retrieve agreement details, metadata, and extracted provisions
Access party information, dates, financial terms, and custom extraction fields
Query agreement status and lifecycle data

eSignature API - envelope operations:

Check envelope status and signer information
List available templates
Create and send envelopes from templates
Retrieve document content and metadata

Workflow Builder API (GA) - workflow operations:

Trigger Workflows programmatically
Pass parameters to workflow instances
Check workflow status

What It Can't Do (Yet)

The MCP server is in Beta, and some capabilities are limited:

No write operations to Agreement Manager - the Agreement Manager API is currently read-only (GET operations). You can search and retrieve agreement data but can't update or annotate agreements through MCP.
No complex document assembly - you can create envelopes from existing templates, but you can't build complex documents with conditional logic through MCP alone.
No Extension App interaction - MCP tools don't directly trigger or interact with Extension App steps within Workflows. The MCP server interacts with the APIs, not the Extension App layer.
Production account access requires approval - Beta access is open for developer accounts, but production deployment requires Docusign approval.
Rate limits apply - the underlying Docusign APIs have rate limits that affect MCP operations. High-frequency AI queries need to be designed with this in mind.

How to Set Up the Docusign MCP Server

Prerequisites

A Docusign developer account (or production account with Beta approval)
An Integration Key (Client ID) and Secret Key from your Docusign Apps and Keys page
An MCP-compatible AI client (Claude Desktop, Cowork, Claude Code, VS Code with Copilot)

Setup for Claude Desktop

Go to your Docusign developer account and navigate to Apps and Keys
Create a new Integration Key and Secret Key
Add the required redirect URIs for the MCP server to your application settings
Open Claude Desktop - Settings - Connectors - Add Custom Connector
Enter the MCP server URL: https://mcp-d.docusign.com/mcp
Enter your Integration Key and Secret Key
Complete the OAuth authentication flow to connect your Docusign account
Claude can now discover and use Docusign tools automatically

Setup for Cowork

Docusign and Anthropic announced a partnership in February 2026 that brings Docusign capabilities directly into Cowork. The setup follows a similar connector pattern - add the Docusign connector through Cowork's settings, authenticate with your Docusign account, and start interacting with your agreements through natural language.

Setup for Claude Code

For developers working in the terminal, Claude Code supports MCP servers natively. Add the Docusign MCP server to your Claude Code configuration, and you can query agreement data, check envelope status, and trigger workflows directly from your development environment.

Alternative: Community MCP Implementations

Beyond the official Docusign MCP server, several community implementations exist:

Luther Systems MCP Server - Python-based, uses JWT authentication for server-to-server integration. Good for headless/automated scenarios where no user interaction is needed.
This Dot Agreement Manager MCP - pre-deployed hosted endpoint focused on Agreement Manager API access. Minimal setup - point your AI client to the hosted URL.

These are useful for specific use cases but aren't officially supported by Docusign.

Security Considerations

Connecting AI agents to your agreement data requires careful security architecture. Agreements contain sensitive business information - financial terms, party details, obligations, and confidential provisions. Here's how to think about security:

Authentication and Access Control

The Docusign MCP server uses OAuth 2.0 authentication. When you connect an AI agent to Docusign through MCP:

User-scoped access - the MCP server operates within the authenticated user's Docusign permissions. If a user can't see an agreement in Docusign, they can't access it through the AI agent either.
OAuth scope limitations - you can control which API capabilities the MCP server can access through OAuth scopes. Read-only scopes for Agreement Manager data are lower risk than scopes that allow envelope creation or workflow triggering.
No credential exposure - the AI agent never sees your Docusign credentials. OAuth tokens are managed by the MCP server layer.

Best Practices

Start with read-only access. Configure the MCP connection with Agreement Manager read access first. Let your team get comfortable with AI-powered agreement search before enabling write operations like envelope sending or workflow triggering.
Review what data the AI can access. The AI agent can search and retrieve any agreement data the authenticated user has access to in Docusign. Make sure your Docusign permission model is correct before connecting MCP.
Audit AI interactions. Docusign's audit trail captures API operations. Log MCP-initiated actions alongside direct user actions for a complete picture of who accessed what.
Use developer accounts for testing. Test your MCP setup thoroughly with a developer account before connecting to production data.
Rotate credentials. Treat MCP Integration Keys like any other API credential - rotate on schedule and revoke if compromised.

What About Sensitive Agreements?

If certain agreements shouldn't be accessible through AI agents (M&A documents, board materials, litigation files), the right approach is Docusign's permission model. Restrict access to those agreements at the Docusign account level, and the MCP server inherits those restrictions automatically. Don't rely on the AI agent to filter sensitive content - control it at the source.

Practical Use Cases

Agreement Search and Intelligence

The highest-value use case today is natural language agreement search through Agreement Manager:

"Find all active NDAs with vendors in healthcare that expire in the next 6 months"
"What's our total committed vendor spend for the next 12 months?"
"Show me all agreements with governing law set to California"
"Which customer agreements have rate escalation clauses?"

These queries would take 15-45 minutes of manual searching. Through MCP, they take seconds.

Envelope Status and Tracking

"What's the status of the MSA we sent to DataTech last week?"
"Which envelopes are waiting on signatures for more than 5 days?"
"How many agreements did we close this month?"

Workflow Initiation

For teams using Workflows:

"Start the client onboarding workflow for Newco Inc - mid-market SaaS, $75K deal, Sarah Chen as primary contact"
"Trigger the vendor compliance review for Q2 renewals"

Agreement Analysis

When Agreement Manager has extracted provision data through Iris AI:

"Compare the liability caps in our agreements with Vendor A versus Vendor B"
"Which of our vendor agreements don't have a data processing addendum?"
"What are the payment terms across our top 10 customer contracts?"

Proactive Monitoring

AI agents can check agreement data on schedule and surface insights:

"You have 3 vendor agreements with auto-renewal opt-out deadlines in the next 30 days"
"Two envelopes sent last week haven't been opened - here are the details"

The Future of AI + Agreements

MCP is the beginning of a broader shift. As AI capabilities evolve and the Docusign MCP server moves from Beta to GA, expect:

Deeper Agreement Manager integration - as the Agreement Manager API expands beyond read-only operations, AI agents will be able to annotate agreements, update metadata, and manage the agreement lifecycle conversationally.

Autonomous agreement processing - AI agents that don't just answer questions but proactively manage agreement lifecycles. They identify expiring contracts, recommend renegotiation strategies, and draft renewal terms based on historical patterns.

Cross-system intelligence - AI agents that correlate agreement data with CRM pipeline, financial data, and project management. "Our agreement with Vendor X expires in 60 days, they've had 3 support escalations this quarter, and there are 2 cheaper alternatives in our vendor assessment pipeline."

Natural language workflow creation - describing what you need in plain language and having an AI agent configure the Workflow, set up triggers, and deploy it.

How Fluidlabs Approaches AI + Agreements

We work with Docusign's APIs daily - building Extension Apps, designing integration architectures, and connecting enterprise systems to the IAM platform. MCP is a natural extension of that work.

What we help with:

MCP strategy - determining which agreement capabilities to expose to AI agents, and in what order, based on your security requirements and use cases
Custom MCP development - for organizations that need capabilities beyond the official Docusign MCP server. Custom MCP servers that combine Docusign data with other business system data for richer AI interactions
Security architecture - designing the access control, audit logging, and data handling policies for AI-mediated agreement access
Integration with Baton - connecting MCP-initiated workflows with Baton's webhook orchestration for end-to-end automation across 40+ platforms

Discuss AI integration for your Docusign environment.

Published by Fluidlabs, Docusign IAM implementation specialists. Get in touch to discuss your implementation.