Signdocs MCP server

SignDocs Brasil — MCP Server

A Model Context Protocol server for the SignDocs Brasil e-signature API. It lets MCP-capable AI clients (Claude Desktop, Claude Code, Cursor, …) create signing sessions, manage multi-signer envelopes, upload/download documents, verify signatures, and manage webhooks — the same action catalog as the official n8n, Zapier, and Make.com integrations.

It is a thin adapter over the official @signdocs-brasil/api SDK, which owns OAuth2 token exchange, caching, retries, and error handling.

Install

npm install -g @signdocs-brasil/mcp-server   # or run on demand with npx

Credentials

Create an API credential in the SignDocs dashboard (app.signdocs.com.br → API) and expose it as environment variables:

Start in hml. HML data expires after ~7 days and is safe for testing.

Switch to production only when you intend to create real, legally-binding

signatures.

Connect an AI client

Claude Desktop (claudedesktopconfig.json):

{
  "mcpServers": {
    "signdocs": {
      "command": "npx",
      "args": ["-y", "@signdocs-brasil/mcp-server"],
      "env": {
        "SIGNDOCS_CLIENT_ID": "your_client_id",
        "SIGNDOCS_CLIENT_SECRET": "your_client_secret",
        "SIGNDOCS_ENVIRONMENT": "hml"
      }
    }
  }
}

Claude Code:

claude mcp add signdocs \
  -e SIGNDOCS_CLIENT_ID=your_client_id \
  -e SIGNDOCS_CLIENT_SECRET=your_client_secret \
  -e SIGNDOCS_ENVIRONMENT=hml \
  -- npx -y @signdocs-brasil/mcp-server

Tools

⚠️ tools carry destructiveHint annotations and a warning in their description so compliant clients prompt the human before invoking them. Annotations are only hints — review your client's auto-approval settings.

Not yet exposed

Trust sessions (/v1/trust-sessions) and resend-invite are not in @signdocs-brasil/api v1.6.1 yet; they'll be added when the SDK supports them. Digital ICP-Brasil A1 signing runs through the lower-level transaction/advance flow rather than a hosted-session profile.

Resources

The server exposes grounding resources the model can read on demand:

Shortened. The full README is on GitHub.