Gorgias

License: MIT Python MCP

MCP server for Gorgias, the e-commerce customer support helpdesk. 5 tools for AI agents to triage tickets, read full conversation context, draft internal notes for human reviewers, send outbound replies when authorized, and look up customer lifetime history.

As of May 2026, Gorgias does not have an official MCP server. The community attempt is a zero-star placeholder. This is the production-quality rail to fill that gap so external agents (Cursor, Claude Desktop, Hermes, LangChain, n8n) can natively manage Gorgias tickets.

The 5 tools

Tool	Purpose
list_open_tickets	Retrieve unresolved support tickets ordered by most-recent update
get_ticket_context	Fetch the entire conversation history of a specific ticket
draft_internal_note	Leave an invisible internal note on a ticket (SAFE path, never visible to customer)
send_ticket_reply	Send an outward-facing reply to the customer (gated by env-flag)
get_customer_history	Look up a customer profile by email for lifetime history grounding

Install

pip install gorgias-mcp

Configure

export GORGIAS_DOMAIN="your-store-subdomain"      # e.g. mystore (NOT mystore.gorgias.com)
export GORGIAS_EMAIL="your-gorgias-user-email"
export GORGIAS_API_KEY="your-gorgias-api-key"
export GORGIAS_ALLOW_OUTBOUND_REPLIES="false"     # set to "true" to enable send_ticket_reply

Get a Gorgias API key in your Gorgias dashboard under Settings -> REST API -> Generate API Key. The auth model is HTTP Basic where username = the Gorgias user email and password = the API key.

Use with Claude Desktop

{
  "mcpServers": {
    "gorgias": {
      "command": "gorgias-mcp",
      "env": {
        "GORGIAS_DOMAIN": "your-store-subdomain",
        "GORGIAS_EMAIL": "your-gorgias-user-email",
        "GORGIAS_API_KEY": "your-gorgias-api-key",
        "GORGIAS_ALLOW_OUTBOUND_REPLIES": "false"
      }
    }
  }
}

Restart Claude Desktop. The 5 Gorgias tools are now available.

Use case: AI customer-support assistant

Typical agent flow:

1. Call list_open_tickets(limit=10) to find recent unresolved tickets
2. For each ticket, call get_ticket_context(ticket_id) to read the full conversation
3. Call get_customer_history(email_address) to ground the response in prior interactions
4. Call draft_internal_note(ticket_id, body_text, sender_id) to leave AI-generated context for the human support agent who will reply
5. (Optional, only if GORGIAS_ALLOW_OUTBOUND_REPLIES=true) Call send_ticket_reply to send an autonomous reply

Safety note

The draft_internal_note tool is the recommended primary path for AI-generated content. Internal notes are invisible to customers and let a human reviewer approve before any outbound action. The send_ticket_reply tool is GATED behind the GORGIAS_ALLOW_OUTBOUND_REPLIES env var (default false). Setting it to true should be deliberate, in workflows where the AI has been explicitly authorized to message customers without review.

Architecture

• Public MIT-licensed wrapper around the Gorgias REST API
• Async HTTP via httpx
• pydantic v2 input validation
• HTTP Basic Auth (base64-encoded GORGIAS_EMAIL:GORGIAS_API_KEY), server-side only
• Outbound-reply gating via env-flag
• Rate-limit aware (429 returns a clean error per Gorgias plan tier)

Development

git clone https://github.com/NoBanks/gorgias-mcp.git
cd gorgias-mcp
pip install -e ".[dev]"
pytest

License

MIT. See LICENSE.

Author

Ryan Hammer (NoBanks). Solo founder + engineer. Built this and 8 other MCP servers as part of a sprint to expose AI agent rails for the products and platforms shipping daily.

• GitHub: @NoBanks
• X/Twitter: @livingagentic
• Site: livingagentic.me, nohumannearby.com

Open to AI engineering roles, contract or full-time, remote-only.