addsign-mcp

MCP server for AddSign — connect Claude (or any MCP client) to your AddSign account so your AI assistant can send documents for e-signature, track signing status, remind pending signers, and download signed PDFs with their audit-trail hashes.

Built and maintained by Gold Bird INC, the maker of AddSign.

Install

Every path needs an AddSign account (free plan included) — get connection instructions tailored to your client, with your API key filled in, at addsign.io/settings/api.

Claude Desktop (one-click, works on the free Claude plan): download the Desktop Extension (.mcpb) and double-click it (or drag it into Claude Desktop). You'll be prompted for your API key — it's stored in your system keychain, never in a file.

claude.ai / Claude Desktop via hosted connector (no key at all): Settings → Connectors → Add custom connector → https://addsign.io/api/mcp, then authorize with your AddSign login (OAuth). Paid Claude plans.

Claude Code:

claude mcp add addsign --env ADDSIGN_API_KEY=sk_your_key -- npx -y addsign-mcp

Any other MCP client (Cursor, VS Code, Windsurf, Gemini CLI, Codex CLI, …):

{
  "mcpServers": {
    "addsign": {
      "command": "npx",
      "args": ["-y", "addsign-mcp"],
      "env": { "ADDSIGN_API_KEY": "sk_your_key" }
    }
  }
}

Tools

Tool	What it does
list_templates	Your templates, their signer roles, and field summaries
send_for_signature	Create a document from a template and email it to signers (supports dry_run previews and idempotent retries)
check_status	Who viewed / signed / declined, with the audit trail
remind	Nudge signers who haven't signed (server-throttled)
download_signed	Short-lived URL for the signed PDF + its ledger SHA-256
list_documents	Your documents, filterable by status (including declined)

Security posture

• The server holds no secrets beyond your own API key, read from the ADDSIGN_API_KEY environment variable only — never from prompts or tool arguments — and redacted from every error it surfaces. It speaks only AddSign's public v1 API; there is no database access.
• No destructive tools. An agent can send, but can never cancel or delete documents, touch account or billing settings, or see API keys. Mis-sends are corrected by the human — every agent-initiated send emails the account owner with one-click View/Cancel.
• Server-side guardrails on every agent send: recipient verification (deliverability, saved-contacts matching, typo detection), optional per-key contacts-only restriction, plan metering, and full audit-ledger attribution of every agent action.
• Keys are created and revoked instantly at addsign.io/settings/api.

Development

npm install
npm run build        # compile TypeScript → dist/
npm run build:mcpb   # build the Claude Desktop Extension → addsign.mcpb

Requires Node.js ≥ 18. The .mcpb build bundles the server into a single file (no node_modules shipped) and packs it with @anthropic-ai/mcpb.

This repository is the public release mirror of the server that ships as addsign-mcp on npm and in the official MCP Registry (io.github.rmindgh/addsign-mcp). Issues and questions are welcome here; the AddSign product itself is not open source.

License

MIT © Gold Bird INC