Skip to main content
Connect DocsAutomator to AI assistants like Claude, Cursor, and other MCP-compatible clients. Generate documents, manage automations, and handle e-signatures directly from your AI conversations.

​
What is MCP?

The Model Context Protocol (MCP) is an open standard that allows AI assistants to securely connect to external tools and data sources. With the DocsAutomator MCP server, you can:
  • Build complete automations: generate Word templates with AI, connect data, map fields, and configure the output
  • Generate documents from your automations
  • List and manage automations
  • Check document generation status
  • View and manage e-signature sessions
  • Extract placeholders from templates, with per-placeholder descriptions for AI clients

​
Quick start: your first quote, accepted

In about two minutes your AI agent makes a quote in your brand and sends it to you for acceptance. You accept it on your phone and see the accepted PDF in DocsAutomator. MCP server URL for other AI apps: https://mcp.docsautomator.co/mcp
1

Add DocsAutomator to your AI app

For Claude and Claude Code, the DocsAutomator skill gives your agent these steps.
2

Paste this prompt

No website? Use this one. Your agent then starts from a quote in the template gallery:
DocsAutomator sends the first document of a new automation in test mode, unless your AI client asks for live. Test mode is free and sends the real email.
3

Accept it on your phone

Open the email, check the quote, type your name and tap Accept. You are done when the quote shows as accepted under Documents, with the accepted PDF.
Next steps. Ask your agent to:
  • send the quote to a real client (the same send without test mode; a real acceptance counts as one document in your plan),
  • make the quote from your Airtable base,
  • use your own Word template.
API key (optional): clients that do not support sign-in can use an API key from DocsAutomator Settings.

​
Multi-Workspace Accounts

If your DocsAutomator account belongs to more than one workspace and you connect through OAuth (Claude, ChatGPT, and other OAuth-based MCP clients), the authorization screen shows a workspace picker. The connection is bound to the workspace you select, and only that workspace’s automations are reachable through it. To change which workspace an existing connection points at, disconnect the integration in your AI client and reconnect. That triggers a fresh OAuth flow and lets you pick a different workspace. API key connections (Cursor, custom clients) are bound to the workspace of the key itself, so no picker is involved; switch workspaces by using a different API key.

​
Setup by Client

​
Claude

DocsAutomator is an official connector in the Claude directory. It works in claude.ai, the Claude desktop app, the Claude mobile app and Claude Code.

Add DocsAutomator to Claude

Opens the DocsAutomator listing in the Claude directory.
  1. Open DocsAutomator in the Claude directory: claude.ai/directory/connectors/docsautomator. You can also go to Customize → Connectors → Discover in Claude and search for DocsAutomator.
  2. Click Connect to Claude.
  3. Sign in to DocsAutomator, choose your workspace if you have more than one, and click Authorize.
  4. Back in Claude, DocsAutomator shows as Connected under Your connectors.
On a Team or Enterprise plan, an owner adds DocsAutomator for the organization first with Connect for your team. Members then click Connect to Claude on the listing and sign in with their own DocsAutomator account. Claude Code: sign in to Claude Code with the same Claude account (/login). DocsAutomator then shows in /mcp as claude.ai DocsAutomator, with no extra setup. If you added DocsAutomator to Claude Code by hand before, remove that entry with claude mcp remove docsautomator, because a server added by hand hides the connector. Once connected, you can ask Claude things like:
  • “List my DocsAutomator automations”
  • “Generate a document using automation [ID] with this data…”
  • “Check the status of my e-signature sessions”

​
ChatGPT

A ChatGPT Business or Enterprise subscription is required to use MCP connections with ChatGPT. Additionally, Developer Mode must be enabled in your ChatGPT settings.
  1. Open ChatGPT and go to Settings
  2. Navigate to Apps
  3. Enable Developer Mode if not already enabled
  4. Click Create App
  5. Select MCP Server as the connection type
  6. Enter the following configuration:
    • Name: DocsAutomator
    • Server URL: https://mcp.docsautomator.co/mcp
    • Authentication: OAuth
    • Leave Client ID and Secret blank
  7. Click Create
Once connected, you can ask ChatGPT to generate documents, manage automations, handle e-signatures, and more using natural language.

​
Cursor

  1. Open Cursor Settings (Cmd/Ctrl + ,)
  2. Navigate to Features → MCP Servers
  3. Click Add Server
  4. Configure:
    • Name: DocsAutomator
    • URL: https://mcp.docsautomator.co/mcp
    • Authentication: Bearer Token
    • Token: Your DocsAutomator API key
Get your API key from DocsAutomator Settings.

​
Other MCP Clients

For MCP-compatible clients other than Claude, use these connection details. Claude connects through the directory connector, with no server URL or API key. A client that does not support sign-in sends your API key in the header Authorization: Bearer YOUR_API_KEY. Get your API key from DocsAutomator Settings.

​
Available Tools

The MCP server gives your AI client 18 tools. Your AI client sees each tool with its full instructions after it connects. The rules DocsAutomator gives your AI client, including the confirmations before any send, are listed in How AI clients use DocsAutomator. The main path: your AI client is the data source. It reads the placeholders with get_automation and sends the values with create_document. Record sources such as Airtable or Notion are an optional second setup.

​
Automations

​
Templates

​
Placeholders

​
Record sources (optional)

​
Output and documents

​
Acceptance and signature

Record source accounts (Airtable, Notion, Google, and the others) use OAuth and are connected once in the DocsAutomator app under Settings → Integrations. When an account is not connected, the tools answer with the link to that page. All other setup steps work through MCP.

​
Descriptions for AI Clients

Two description fields help an AI client work with your automations without extra prompt engineering:
  • Automation description: a note on what the automation is for and what data it needs. Set it in the app under Automation Settings → Description, or with configure_automation. list_automations returns it, so the AI picks the right automation.
  • Placeholder descriptions: what each placeholder means and how it prints (for example “Subtotal before tax, in EUR; prints as 1.234,56 €”). Set them with the gear icon next to a mapping, or with configure_placeholders. get_automation returns them (the REST API too), so the AI knows what data belongs in each placeholder.
DocsAutomator formats every value itself, so an AI client sends raw values: numbers as numbers and dates as ISO dates (2026-10-01). Example prompt:
“Describe the placeholders of automation abc123 so my CRM agent can fill them.”

​
Make a document

create_document parameters:
  • automationId (required): the automation ID
  • mode: preview (a watermarked PDF, nothing sent), test (outputs run in test mode) or live (a real document)
  • data: placeholder values, sent raw
  • recId: the record ID for Airtable, Notion, SmartSuite and ClickUp automations
  • rowNumber: the row number for Google Sheets automations
  • documentName: a custom name for the document
  • async: if true, answers at once with a job ID; follow it with get_recent_runs
  • recipients: who signs or accepts this document, for example [{ "signer": 1, "email": "anna@acme.com", "name": "Anna Weber" }]
  • skipEsign: if true, makes the document without starting signing or acceptance
Signing and acceptance: if the automation has signing or acceptance turned on, create_document starts it. The answer includes the session and its signing links. Recipients named in chat: the assistant always passes recipients when a document goes out for signing or acceptance. When you tell the assistant who should sign or accept (“send quote 12 to Anna at anna@acme.com”), it passes that person in recipients. This works for every data source, including Airtable, Notion, Google Sheets, SmartSuite and ClickUp automations called with a record ID, and you do not need a second automation to send from chat. A listed signer’s email and name replace the saved ones for that one document; the automation itself does not change. signer is the signer’s number from get_automation. It can be left out when the automation has exactly one signer and is needed when it has two or more. When the assistant sets up signing or acceptance with configure_esignature, it follows these defaults:
  • On API and MCP automations, and automations without a data source, every signer is “given with each request”, so each send names the person in recipients.
  • On Zapier, Make, n8n, Glide and Noloco automations, the signer uses a template placeholder, because their apps do not send recipients with each run.
  • On Airtable, Notion, Google Sheets, SmartSuite, ClickUp and Google Forms automations, the signer points at the record’s email field, because their triggers run without anyone naming a recipient.
Test mode: DocsAutomator sends the first document of a new automation in test mode when your AI client does not ask for a mode, and it refuses a live document from an automation that has not made a test document yet. A test mode signing or acceptance request sends the real email to the real person and is never billed. Confirmation before a send: before anything goes out by email, DocsAutomator can ask you in a confirmation dialog that names the document and every recipient. Nothing is sent until you confirm, and a preview never asks. How AI clients use DocsAutomator lists when it asks and which AI clients show the dialog. Example prompts:
“Create a document using automation abc123 for Acme Corp, dated today.” “Send the quote for record rec123 to Anna Weber at anna@acme.com for acceptance.” “Who has not accepted their quote yet? Remind them.”

​
E-invoices

When the e-invoice add-on is on for the workspace and the automation, a document that Claude, ChatGPT or another MCP client creates with create_document is also an e-invoice. Nothing changes in the conversation; the automation’s E-invoice section decides the format and the fields.
  • Mode test or live: with ZUGFeRD, the pdfUrl the client gets is the e-invoice. With XRechnung, the XML goes to the automation’s XML field, Google Drive and email as set up; the answer to the client carries only the PDF link, not the XML link.
  • Mode preview: the client gets a watermarked PDF. If the data would not make a valid e-invoice, the preview shows the plain PDF and does not report the problems. Use test to check an e-invoice from a chat, or the Test document card in the app.
  • Problems: if the data cannot make a valid e-invoice, no document is created, and the client receives the message that names each problem (for example “Gross total does not match the line items”), so it can fix the data and try again.
  • Values: send amounts as plain euro numbers and dates as ISO dates, as for every document.

​
Example Conversations

​
Generate a Contract

You: “I need to generate a consulting agreement for Acme Corp. Use the consulting agreement automation.” Claude: “I’ll create that document for you. Let me first check what placeholders are needed…” [Uses get_automation, then create_document] Claude: “Done! Here’s your consulting agreement:
  • PDF: [link]
  • Google Doc: [link]
The e-signature request has been sent to the signers automatically.”

​
Check Signing Status

You: “What’s the status of the contract I sent to john@acme.com?” Claude: [Uses list_esign_sessions with email filter] “I found one pending session for john@acme.com:
  • Document: Consulting Agreement - Acme Corp
  • Status: Waiting for signature
  • Sent: 2 hours ago
Would you like me to resend the invitation?”

​
Batch Document Generation

You: “Generate invoices for all the rows in my Google Sheet automation” Claude: “I’ll generate those invoices. Since there are multiple documents, I’ll use async mode…” [Uses create_document with async=true for each row, then get_recent_runs for each job]

​
Troubleshooting

​
”Authentication required” error

  • Ensure your API key is valid and active
  • Check that the Authorization header is formatted correctly: Bearer YOUR_API_KEY
  • Verify your API key at DocsAutomator Settings

​
”Automation not found” error

  • Verify the automation ID is correct
  • Ensure the automation belongs to your workspace
  • Check that the automation is not deleted

​
E-signature not starting

  • Verify the automation has e-signing enabled in its output settings
  • Check that signer emails are configured in the automation
  • Ensure the e-signature placeholders exist in the template

​
Rate limiting

  • Document generation: 2000 requests per 15 minutes
  • If you hit limits, wait for the reset period
Last modified on October 13, 2026