Give Your AI Agent E-Signature Powers: Signbee MCP Server Guide (2026)

Why AI agents need e-signature capabilities

Most developers use AI agents to automate software workflows: writing code, running tests, or updating documentation. However, when an autonomous sales agent or onboarding agent reaches the “signature phase,” they hit a wall. Traditional e-signature providers make programmatic interaction unnecessarily difficult. DocuSign, for instance, has hundreds of endpoints and expects complex OAuth configurations that are impossible for an autonomous LLM to navigate.

The Model Context Protocol (MCP) solves this. By exposing Signbee's clean, markdown-based signing endpoint as a direct MCP tool, your AI assistant can write a contract, populate variables from workspace context, and trigger a legally binding signature cycle. No template builders, no iframe embed code, and no manual copy-pasting required.

Configuring signbee-mcp: Cursor, Claude Desktop, and Windsurf

Connecting the Signbee MCP server to your desktop client takes less than 2 minutes. The server is packaged as an npm package (signbee-mcp) and runs via npx.

1. Claude Desktop Setup

Claude Desktop reads its custom tools from a local configuration file. Open this file on your machine:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Paste the following configuration block inside the mcpServers object:

{
  "mcpServers": {
    "signbee": {
      "command": "npx",
      "args": ["-y", "signbee-mcp"],
      "env": {
        "SIGNBEE_API_KEY": "your_api_key_from_dashboard"
      }
    }
  }
}

Restart Claude Desktop. A small hammer icon will appear in the chat bar, showing that the Signbee signing tools are active.

2. Cursor Setup

Cursor provides two ways to configure MCP. You can add it directly via the GUI settings (Settings > Features > MCP), or you can define it globally or on a project-level by creating a .cursor/mcp.json file in your workspace:

{
  "mcpServers": {
    "signbee": {
      "command": "npx",
      "args": ["-y", "signbee-mcp"],
      "env": {
        "SIGNBEE_API_KEY": "your_api_key_from_dashboard"
      }
    }
  }
}

3. Windsurf Setup

For Windsurf's Cascade agent, the config file is located in the user directory:

• macOS/Linux: ~/.codeium/windsurf/mcp_config.json
• Windows: %USERPROFILE%\.codeium\windsurf\mcp_config.json

Add the same Signbee server block to the file and restart Windsurf to reload the cascade panel.

Prompting patterns for document-generating agents

Once the MCP tools are active, you can query your agent using natural language. To get the best results, use prompts that provide explicit recipient details, scope, and values.

Because Signbee processes Markdown directly into PDF on the fly, the agent has complete control over content layout. It can format sections, inject tables, and style bold clauses based on the exact parameters in your chat context.

Autonomous drafting and signing: under the hood

When an LLM triggers the e-signing flow, the signbee-mcp client runs a standard fetch command under the hood. If you are developing custom autonomous agents outside Cursor/Claude, you can recreate this exact behavior using Python or Node.js.

Here is a Python example using httpx to query the status of a document programmatically in a background task:

import httpx
import asyncio

async def check_contract_status(document_id: str, api_key: str):
    # Query Signbee API for the live status of the agreement
    url = f"https://signb.ee/api/v1/documents/{document_id}"
    headers = {
        "Authorization": f"Bearer {api_key}",
        "Accept": "application/json"
    }
    
    async with httpx.AsyncClient() as client:
        response = await client.get(url, headers=headers)
        if response.status_code == 200:
            data = response.json()
            status = data.get("status")
            print(f"Document ID: {document_id} | Status: {status}")
            
            if status == "signed":
                print(f"Cryptographic Hash: {data.get('sha256_hash')}")
                print(f"Signed At: {data.get('signed_at')}")
            return data
        else:
            print(f"Failed to check status. Code: {response.status_code}")
            return None

# Example execution
# asyncio.run(check_contract_status("doc_123456", "your-api-key"))

Ensuring safety with human-in-the-loop (HITL) review

Allowing an LLM to generate and send legal documents autonomously raises safety concerns. A minor prompting hallucination could result in incorrect rates, incorrect client names, or unintended liability clauses being delivered directly to a client.

To prevent this, you should enforce a Human-in-the-Loop (HITL) strategy using one of three patterns:

API key authentication vs OTP verification

The Signbee MCP server supports two modes of verification depending on whether an API key is present:

For local testing in Cursor, you can safely skip the API key env variable. The recipient gets the document exactly the same way, but you will receive an OTP mail to verify your sender identity. For production deployments and automated agent scripts, the API key is required to keep operations running smoothly without manual friction.

Frequently Asked Questions

Last updated: May 29, 2026 · Michael Beckett is the founder of Signbee and B2bee Ltd.