
Your Agents Can Now Act on Agreements: The Docusign MCP Server Is Generally Available
The Docusign MCP Server is now generally available. Learn how to connect AI agents to Docusign to send, track, search, and analyze agreements, explore new capabilities, and get started.
The Docusign MCP Server is now generally available, providing a standardized way for AI agents to send, track, search, and analyze agreements. It also serves as the underlying platform powering Docusign's growing ecosystem of AI connectors, including Claude, ChatGPT, Gemini, and Copilot.
If you're already building on Docusign APIs, your Integration Keys (IK), OAuth authentication flows, and permission scopes remain unchanged. Instead of writing custom orchestration code and API callers for every workflow, you can expose specific Docusign MCP tools to your agent, configure human-in-the-loop approval gates, and let the model execute the operational steps.
For new integrations, the MCP Server removes the need to implement the full API schema upfront. Create a developer account, connect your MCP-compliant client, and start making tool calls. The Docusign APIs remain available when you need more granular control.
What’s new since Open Beta
Since Open Beta, we’ve expanded where the Docusign connector is available and added key capabilities to better support enterprise customers. This includes support for Perplexity and Slackbot, along with an in-chat upload widget in Claude that lets users send up to 10 local documents for signature.
New tools let agents retrieve agreements and contract details, including CLM documents, and list, start, or stop CLM workflows. You can now send envelopes for signing with documents uploaded from your local device using an embedded UI experience inside AI assistants. Envelope searches now return twice as many results per call, agreement searches flag when results are partial, and admins can turn MCP access on or off for each account.
Get started with the Docusign MCP Server
Create a free Docusign developer account
In Apps and Keys, create an IK with a secret key and your client’s redirect URIs
In your AI client, add a custom connector pointing to https://mcp-d.docusign.com/mcp
Sign in and try: “Which of my envelopes are waiting on a signer, and who’s holding them up?”
Explore the tool reference and authentication guide in the Docusign MCP Server documentation, and bring questions to the MCP thread in the Docusign Community.
What can you do with the Docusign MCP Server?
The server exposes tools that let external agents perform agreement-related tasks across eSignature, Agreement Manager, Workflow Builder, and select Contract Lifecycle Management (CLM) capabilities.
External agents can:
Prepare and send envelopes from templates or from documents uploaded from a local device
Track envelope and recipient status
List, trigger, and pause Workflow Builder workflows
List, start, and stop CLM workflows
Query agreements by standard provisions, such as governing law or expiration date, and by custom provisions your organization configured in Agreement Manager
For customers with eligible CLM and IAM plans, agreement queries include CLM documents from before and after execution. The tools available depend on the user’s plan.
In Claude and ChatGPT, an MCP App upload widget lets users send local documents for signature without a template. MCP App interfaces are currently limited to Claude and ChatGPT, with other clients and custom agents to follow.
Who should use the Docusign MCP Server?
The Docusign MCP Server supports a range of builders, from engineers developing custom agents to operations leads creating automations without code. That includes teams at startups and global enterprises, as well as the partners who serve them.
Most builders fall into three groups:
In-house builders, creating agents for their own company
Partners and product teams, building agents for their customers
Hands-on individuals, automating their own day-to-day work
Here’s what each group can build and how to get started.
In-house builders: Agents for your own company
For example, a Copilot Studio agent could answer, “Do we already have an NDA with this vendor, and when does it expire?” before procurement sends a new one. A contract-request agent built in your own code could start the right CLM workflow as soon as a request comes in, rather than leaving it in legal’s inbox.
In AI platforms, build on the official connector. For custom agents, use your own allowlisted IK and build the approval step yourself.
Want to simplify further and to run your agentic workflow on a schedule? Try a scheduled job or an MCP-capable automation platform like n8n, starting in the demo environment. A weekly check for agreements nearing auto-renewal is a good first job. Use JWT Grant: tokens expire after eight hours, so create refresh logic to let these automations run unattended.
Start here: build against a developer account with a custom connector.
Partners and product teams: Agents for your customers
A property-management platform could let its agent send leases from each customer’s own Docusign templates and report who has signed. A systems integrator (SI) could build a renewal desk for a client’s legal team that tracks expiration dates and the client’s custom provisions.
Your agent uses your allowlisted IK. Each customer signs in with their own account, and their admin enables MCP, allowing one agent to serve many customers.
Start here: Prototype one client workflow in demo, then request allowlisting.
No-code builders: Automate agreement tasks without writing code
Operations leads and power users in legal, sales, procurement, and HR can use the official Docusign connector in Claude to automate recurring agreement tasks without writing code or using an Integration Key (IK).
For example, a legal ops lead could create a skill that sends NDAs from an approved template and schedule a weekly summary of agreements nearing auto-renewal. A sales ops manager could set up a daily report of envelopes that have been waiting on a signer for more than three days. These tasks can also require approval before execution.
Start with: Connect the official Docusign connector in Claude and create a scheduled task using a prompt like:
“Use the Docusign connector to list envelopes I’ve sent that have been waiting on a signer for more than three days, with each recipient’s name and the date sent.”
Run the prompt once in chat to check the results, then schedule it to run every Monday at 9 a.m.
Build in demo, ship in production
The Docusign MCP Server offers separate demo and production endpoints with the same tools, so you can test your integration before connecting to live accounts.
Demo (
https://mcp-d.docusign.com/mcp) connects to a free developer account with test data, so you can test tools and edge cases like voids without touching real agreements.Production (
https://mcp.docusign.com/mcp) connects to live accounts and real agreements.
How you connect depends on where you're building. Official Docusign connectors only connect to production accounts, so you'll need to add a custom connector with your own Integration Key (IK) to work in demo. Custom connectors are self-managed and may receive new tools first.
Where you build | Demo (build and test) | Production (go live) |
AI assistants with an official connector, such as Claude, ChatGPT, Gemini, Perplexity, Slack, and Copilot Studio | Custom connector with your IK | Official Docusign connector |
Your own agent or platform | Your IK | Your allowlisted IK |
For AI assistants in production, we recommend using the official Docusign connector, which Docusign maintains and each platform partner reviews and monitors. You can point a custom connector at the production endpoint, but this isn't recommended.
If you're building your own agent or platform, submit the allowlist request form opens in a new tabbefore using your IK in production. An in-product attestation step is also coming to the Developer Console.
Security, governance, and data residency
The MCP Server includes controls for authentication, permissions, and administration, along with safeguards for how agents access and use agreement data.
Your permissions, nothing more. The stateless server acts only for the authenticated user, so an agent can’t access anything the user can’t.
Admin control. Admins can turn MCP access on or off per account under Admin > Integrations > MCP Connections.
Human approval. Claude, ChatGPT, and other AI assistants prompt users before actions run. Custom agents and scheduled jobs must add their own approval step.
Data residency. The server runs in Docusign’s regional hosting: US, EU, Australia, Canada, and Japan. Once data is returned to an AI client, that client’s data handling terms apply.
Prompt injection. Free-text agreement fields are marked as untrusted in tool schemas, helping agents treat them as data rather than instructions.
Language support. The MCP Server and MCP Apps interface are released in English only. At GA, non-English interactions are handled by the connected third-party LLM, so response quality depends on the AI assistant’s language capabilities.
Note: The server isn’t yet available for Docusign Government (FedRAMP) accounts. For details, see the security documentation.
Additional resources
For more information on building with the Docusign MCP Server, explore these resources:
Docusign MCP Server documentation: Setup instructions, authentication, and tool reference
Docusign MCP Server security documentation: Security, governance, and supported environments
Docusign Community MCP thread: For questions and discussion with other builders
MCP Server use cases and examples: Explore practical use cases and ideas for building with the MCP Server

Gaurav Rajoria is a Lead Product Manager in Docusign, driving the product vision and strategy for the Docusign MCP Server. His focus is to bring Docusign capabilities into users’ AI tools of choice and to empower developers to integrate agreement scenarios seamlessly into their AI agents. You can reach out to Gaurav on Linkedin.
Related posts
Docusign IAM is the agreement platform your business needs




