SignNow MCP Server

A Model Context Protocol (MCP) server that gives AI agents secure, structured access to SignNow eSignature workflows — templates, embedded signing, invites, status tracking, and document downloads — over STDIO or Streamable HTTP.

mcp-name: io.github.signnow/sn-mcp-server

Table of contents

• Features

• Quick start

  • Prerequisites
  • Quick run (uvx)
  • 1. Setup Environment Variables
  • 2. Install and Run
  • Local/Remote (HTTP)
  • Docker
  • Docker Compose

• Configuration

  • Authentication options
  • SignNow & OAuth settings
  • Production key management

• Client setup

  • VS Code — GitHub Copilot (Agent Mode) / Cursor
  • Claude Desktop
  • Glama (hosted MCP)
  • MCP Inspector (testing)

• Tools

• FAQ / tips

• Examples

• Useful resources

  • Sample apps
  • API documentation
  • SignNow API Helper MCP

• License

Features

• Templates & groups

  • Browse all templates and template groups
  • Create documents or groups from templates (one-shot flows included)

• Invites & embedded UX

  • Email invites and ordered recipients
  • Embedded signing/sending/editor links for in-app experiences

• Status & retrieval

  • Check invite status and step details
  • Download final documents (single or merged)
  • Read normalized document/group structure for programmatic decisions

• Transports

  • STDIO (best for local clients)
  • Streamable HTTP (best for Docker/remote)

Quick start

Prerequisites

• SignNow account. Create a free developer account.
• SignNow Credentials: You will need your account email, password, and the application Basic Authorization Token. Getting started.
• An active SignNow API application.
• Python 3.11+ installed on your system (check with python3 --version)
• UVX installed  (check with uvx --version). Recommended for the quickest setup.
• Environment variables configured
• If your client supports Streamable HTTP, you can use the pre-deployed server URL https://mcp-server.signnow.com/mcp instead of running it locally.

Quick run (uvx)

If you use uv, you can run the server without installing the package:

uvx --from signnow-mcp-server sn-mcp serve

1. Setup Environment Variables

# Create .env file with your SignNow credentials
# You can copy from env.example if you have the source code
# Or create .env file manually with required variables (see Environment Variables section below)

2. Install and Run

Option A: Install from PyPI (Recommended)

# Install the package from PyPI
pip install signnow-mcp-server

# Run MCP server in standalone mode
sn-mcp serve

Option B: Install from Source (Development)

# 1) Clone & configure
git clone https://github.com/signnow/sn-mcp-server.git
cd sn-mcp-server
cp .env.example .env
# fill in your values in .env

# 2) Install (editable for dev)
pip install -e .

# 3) Run as STDIO MCP server (recommended for local tools & Inspector)
sn-mcp serve

STDIO is ideal for desktop clients and local testing.

Local/Remote (HTTP)

# Start HTTP server on 127.0.0.1:8000
sn-mcp http

# Custom host/port
sn-mcp http --host 0.0.0.0 --port 8000

# Dev reload
sn-mcp http --reload

By default, the Streamable HTTP MCP endpoint is served under /mcp. Example URL:

http://localhost:8000/mcp

Docker

# Build
docker build -t sn-mcp-server .

# Run HTTP mode (recommended for containers)
docker run --env-file .env -p 8000:8000 sn-mcp-server sn-mcp http --host 0.0.0.0 --port 8000

STDIO inside containers is unreliable with many clients. Prefer HTTP when using Docker.

Docker Compose

# Only the MCP server
docker-compose up sn-mcp-server

# Both services (if defined)
docker-compose up

Configuration

Copy .env.example → .env and fill in values. All settings are validated via pydantic-settings at startup.

Authentication options

1) API Key / Access Token (simplest)

SIGNNOW_ACCESS_TOKEN=<your_api_key>
# or equivalently:
SIGNNOW_API_KEY=<your_api_key>

Generate an API key →

2) Username / Password

SIGNNOW_USER_EMAIL=<email>
SIGNNOW_PASSWORD=<password>
SIGNNOW_API_BASIC_TOKEN=<base64 basic token>

3) OAuth 2.0 (for hosted/advanced scenarios)

SIGNNOW_CLIENT_ID=<client_id>
SIGNNOW_CLIENT_SECRET=<client_secret>
# + OAuth server & RSA settings below

When running via some desktop clients, only user/password may be supported.

Per-request access token (HTTP transport)

For multi-tenant / proxy deployments, a wrapping backend can attach the raw SignNow access token per request via a dedicated HTTP header — no env config, and without reusing Authorization (which carries the MCP/OAuth bearer):

X-SignNow-Access-Token: <raw SignNow access_token>

• Value is the raw token — no Bearer prefix.
• Resolved per request; nothing is cached or stored (stateless).
• Precedence: an env-configured SIGNNOW_ACCESS_TOKEN (Option 1) still wins. To use the header as a true per-request credential, do not also set SIGNNOW_ACCESS_TOKEN — otherwise every request collapses to the env token. The header outranks a generic Authorization: Bearer.
• HTTP transport only (sn-mcp http). It is not exposed as a tool argument, so the token never enters the model's context or conversation logs. Send over HTTPS.

SignNow & OAuth settings

# SignNow endpoints (defaults shown)
SIGNNOW_APP_BASE=https://app.signnow.com
SIGNNOW_API_BASE=https://api.signnow.com

# OAuth server (if you enable OAuth mode)
OAUTH_ISSUER=<your_issuer_url>
ACCESS_TTL=3600
REFRESH_TTL=2592000
ALLOWED_REDIRECTS=<comma,separated,uris>

# RSA keys for OAuth (critical in production)
OAUTH_RSA_PRIVATE_PEM=<PEM content>
OAUTH_JWK_KID=<key id>

Production key management

If OAUTH_RSA_PRIVATE_PEM is missing in production, a new RSA key will be generated on each restart, invalidating all existing tokens. Always provide a persistent private key via secrets management in prod.

Client setup

VS Code — GitHub Copilot (Agent Mode) / Cursor

Create .vscode/mcp.json / .cursor/mcp.json in your workspace:

STDIO (local):

{
  "servers": {
    "signnow": {
      "command": "sn-mcp",
      "args": ["serve"],
      "env": {
        "SIGNNOW_USER_EMAIL": "${env:SIGNNOW_USER_EMAIL}",
        "SIGNNOW_PASSWORD": "${env:SIGNNOW_PASSWORD}",
        "SIGNNOW_API_BASIC_TOKEN": "${env:SIGNNOW_API_BASIC_TOKEN}"
      }
    }
  }
}

STDIO (uvx — no local install):

{
  "servers": {
    "signnow": {
      "command": "uvx",
      "args": ["--from", "signnow-mcp-server", "sn-mcp", "serve"],
      "env": {
        "SIGNNOW_USER_EMAIL": "${env:SIGNNOW_USER_EMAIL}",
        "SIGNNOW_PASSWORD": "${env:SIGNNOW_PASSWORD}",
        "SIGNNOW_API_BASIC_TOKEN": "${env:SIGNNOW_API_BASIC_TOKEN}"
      }
    }
  }
}

HTTP (remote or Docker):

{
  "servers": {
    "signnow": {
      "type": "http",
      "url": "http://localhost:8000/mcp"
    }
  }
}

Then open Chat → Agent mode, enable the signnow tools, and use them in prompts.

Note: The same configuration applies in Cursor — add it under MCP settings (STDIO or HTTP). For STDIO, you can also use uvx as shown above.

Claude Desktop

Use Desktop Extensions or the manual MCP config (Developer → Edit config).

Steps:

1. Open Claude Desktop → Developer → Edit config
2. Add a new server entry under mcpServers
3. Save and restart Claude Desktop

Examples:

STDIO (local install):

{
  "mcpServers": {
    "signnow": {
      "command": "sn-mcp",
      "args": ["serve"],
      "env": {
        "SIGNNOW_USER_EMAIL": "${env:SIGNNOW_USER_EMAIL}",
        "SIGNNOW_PASSWORD": "${env:SIGNNOW_PASSWORD}",
        "SIGNNOW_API_BASIC_TOKEN": "${env:SIGNNOW_API_BASIC_TOKEN}"
      }
    }
  }
}

STDIO (uvx — no local install):

{
  "mcpServers": {
    "signnow": {
      "command": "uvx",
      "args": ["--from", "signnow-mcp-server", "sn-mcp", "serve"],
      "env": {
        "SIGNNOW_USER_EMAIL": "${env:SIGNNOW_USER_EMAIL}",
        "SIGNNOW_PASSWORD": "${env:SIGNNOW_PASSWORD}",
        "SIGNNOW_API_BASIC_TOKEN": "${env:SIGNNOW_API_BASIC_TOKEN}"
      }
    }
  }
}

HTTP (remote or Docker):

{
  "mcpServers": {
    "signnow": {
      "type": "http",
      "url": "http://localhost:8000/mcp"
    }
  }
}

Then enable the server in Claude’s chat and start using the tools.

Glama (hosted MCP)

Deploy and run this server on Glama with minimal setup:

Steps:

1. Open the server page on Glama: sn-mcp-server on Glama
2. Click the red "Deploy Server" button
3. In environment variables, provide:
   • SIGNNOW_USER_EMAIL
   • SIGNNOW_PASSWORD
   • SIGNNOW_API_BASIC_TOKEN
   • (other variables can be left as defaults)
4. Create an access token in Glama and copy the endpoint URL. It will look like:

https://glama.ai/endpoints/{someId}/mcp?token={glama-mcp-token}

Use this HTTP MCP URL in any client that supports HTTP transport (e.g., VS Code/Cursor JSON config or Claude Desktop HTTP example above).

MCP Inspector (testing)

Great for exploring tools & schemas visually.

# Start Inspector (opens UI on localhost)
npx @modelcontextprotocol/inspector

# Connect (STDIO): run your server locally and attach
sn-mcp serve

# Or connect (HTTP): use http://localhost:8000/mcp

You can list tools, call them with JSON args, and inspect responses.

Tools

Each tool is described concisely; use an MCP client (e.g., Inspector) to view exact JSON schemas.

• list_all_templates — List templates & template groups with simplified metadata. Supports limit/offset pagination (default: 50 items per page).
• list_contacts — Search CRM contacts by name, email, or phone. Returns id, email, first/last name, and company. Use before send_invite to resolve a recipient's email address by their name. Supports per_page (default 15, max 100).
• list_documents — Browse your documents, document groups and statuses. Supports limit/offset pagination (default: 50 items per page).
• create_from_template — Make a document or a group from a template/group.
• create_template — Convert a document or document group into a reusable template for signing.
• send_invite — Email invites (documents or groups), ordered recipients supported. Auto-detects freeform documents (no fields) — omit role for freeform recipients. Pass self_sign=True (with no orders) to sign the document yourself: the tool resolves your email server-side and returns a SendInviteResponse whose optional link field holds a ready-to-open signing link (also populated when a freeform recipient email matches the authenticated user). For template/template group it auto-creates document/document group first.
• create_embedded_invite — Embedded signing session without email delivery for documents/groups/templates. For template/template group it auto-creates document/document group first.
• create_embedded_sending — Embedded “sending/management” experience for documents/groups/templates. For template/template group it auto-creates document/document group first.
• create_embedded_editor — Embedded editor link to place/adjust fields for documents/groups/templates. For template/template group it auto-creates document/document group first.
• send_invite_from_template — One-shot: create from template and invite.
• create_embedded_sending_from_template — One-shot: template → embedded sending.
• create_embedded_editor_from_template — One-shot: template → embedded editor.
• create_embedded_invite_from_template — One-shot: template → embedded signing.
• get_invite_status — Current invite status/steps for document or group. Covers field invites and freeform invites (field path preferred when both exist). Response includes invite_mode (field or freeform). For freeform document groups, signer emails come from the group documents list (signature_requests).
• get_document_download_link — Direct download link (merged output for groups).
• get_signing_link — Get signing link for a document or document group.
• get_document — Normalized document/group structure with field values, plus the folder the entity is stored in (folder_id/folder_name) at v3.0.
• update_document_fields — Prefill text fields in individual documents.
• upload_document — Upload a file from a local file path (file_path), public URL (file_url), or MCP resource attachment (resource_uri). Set kind='template' (default 'document') to store it as a reusable template instead of a regular document (the returned document_id is then a template ID and next_steps switch to the template follow-ups). For file_path, the resolved path must stay within the configured safe base directory (by default, the user's home directory); paths outside that base fail validation. Supported: PDF, DOC, DOCX, PNG, JPG, JPEG. Max 40 MB. Returns document_id, filename, source.
• send_invite_reminder — Send a signing reminder to pending signers on a document or document group.
• cancel_invite — Cancel all active (pending) signing invites on a document or document group. Auto-detects entity type and invite type (field vs freeform). Returns status: cancelled, completed (already done), or invite_not_sent.
• update_invite_recipient — Replace the signing recipient on a pending field invite. Finds the pending invite for the current signer and swaps in a new email. Supports both documents and document groups. Only field invites — freeform/embedded are unsupported.
• view_document — Generate a read-only embedded view link for a document or document group. In MCP Apps-compatible clients the document renders inline; in other hosts the link is returned as a clickable URL.
• rename_entity — Rename a document, document group, template, or template group. Auto-detects entity type when not provided.
• signnow_skills — Query the bundled SignNow skill library. Omit skill_name to list all available skills with descriptions; provide skill_name (e.g. signnow101) to fetch the full Markdown body. Use signnow101 to learn SignNow entity types, invite types, and tool mappings.
  • List mode example: {"skills": [{"name": "signnow101", "description": "SignNow 101 concepts reference... (description truncated for brevity)"}]}
  • Fetch mode example: {"name": "signnow101", "body": "# SignNow 101 — Concepts Reference\n..."}

Tip: Start with signnow_skills (no arguments) to discover available skills, then list_all_templates → create_from_template → create_embedded_* / send_invite, then get_invite_status and get_document_download_link.

FAQ / tips

• STDIO vs Docker? Prefer STDIO for local dev; inside Docker, use HTTP.
• Sandbox vs production? Start with SignNow’s sandbox/dev credentials; production requires proper OAuth and persistent RSA private key.
• Where do I see exact tool schemas? Use MCP Inspector or your client’s “tool details” view.
• Where are examples? See examples/ in this repo for starter integrations.

Examples

The examples/ directory contains working examples of how to integrate the SignNow MCP Server with popular AI agent frameworks:

• LangChain - Integration with LangChain agents using langchain-mcp-adapters
• LlamaIndex - Integration with LlamaIndex agents using llama-index-tools-mcp
• SmolAgents - Integration with SmolAgents framework using native MCP support

Each example demonstrates how to:

• Start the MCP server as a subprocess
• Convert MCP tools to framework-specific tool formats
• Create agents that can use SignNow functionality
• Handle environment variable configuration

To run an example:

# Make sure you have the required dependencies installed
pip install langchain-openai langchain-mcp-adapters  # for LangChain example
pip install llama-index-tools-mcp                   # for LlamaIndex example  
pip install smolagents                              # for SmolAgents example

# Set up your .env file with SignNow credentials and LLM configuration
# Then run the example
python examples/langchain/langchain_example.py
python examples/llamaindex/llamaindex_example.py
python examples/smolagents/stdio_demo.py

Useful resources

Sample apps

Explore ready-to-use sample apps to quickly test preparing, signing, and sending documents from your software using the SignNow API.

Try the sample apps.

API documentation

Find technical details on SignNow API requests, parameters, code examples, and possible errors. Learn more about the API functionality in detailed guides and use cases.

Read the API documentation.

SignNow API Helper MCP

Connect your AI to access API docs, generate code for complex signing workflows, and troubleshoot integration errors automatically. Access the API Helper MCP

License

MIT — see LICENSE.md.

About SignNow MCP Server — maintained by the SignNow team. Issues and contributions welcome via GitHub pull requests.