Skip to main content
Glama

CFT MCP Server

Remote MCP server for the Chrome Fair Trade platform, intended for deployment on Render.

What was fixed

  • Renamed package-1.json to package.json

  • Added a real TypeScript build pipeline

  • Fixed the start script to run compiled output from dist/index.js

  • Fixed tsconfig.json so TypeScript actually includes src/**/*.ts

  • Added required dev dependencies: typescript, tsx, @types/node, @types/express

  • Removed the hard-coded fallback admin secret from runtime code

  • Updated render.yaml so CFT_ADMIN_SECRET is not committed in plaintext

  • Added a /health route and safer MCP endpoint error handling

Related MCP server: WAHA MCP Server

Tools

Tool

What it does

cft_health

Platform + DB health check

cft_market_prices

Live CIF China + FOT rates, all grades

cft_list_deals

All deals, filter by status

cft_deal_stats

KPIs: tonnage, fees, grade breakdown

cft_whatsapp_sessions

Active WhatsApp conversation states

cft_reset_whatsapp_session

Unstick a supplier's WhatsApp flow

cft_update_deal_status

Approve / reject a deal

cft_verify_certificate

QR-verified deal certificate lookup

cft_calculate_fot

Reverse-Netback FOT calculation

cft_list_suppliers

All registered supplier accounts

Local development

npm install
npm run build
CFT_BASE_URL=https://cft-terminal-hub.vercel.app CFT_ADMIN_SECRET=your-secret npm start

Server routes:

  • GET /

  • GET /health

  • POST /mcp

Render deployment

Render accepts custom build and start commands for Node services, which fits this project layout. citeturn150812view3

1. Push to GitHub

git init
git add .
git commit -m "fix: productionize CFT MCP server"
git branch -M main
git remote add origin https://github.com/Afrinov8/cft-mcp-server.git
git push -u origin main

2. Create the service on Render

  • New → Web Service

  • Connect the repository

  • Render will use:

    • Build Command: npm install && npm run build

    • Start Command: npm start

3. Set environment variables

Set these in Render:

  • CFT_BASE_URL=https://cft-terminal-hub.vercel.app

  • CFT_ADMIN_SECRET=...your actual admin secret...

  • NODE_ENV=production

Do not commit the real admin secret into the repo. Render blueprints support sync: false for secrets so values are not hardcoded in render.yaml. citeturn652236search0turn652236search3

4. Verify after deploy

  • Root health: https://your-service.onrender.com/health

  • MCP endpoint: https://your-service.onrender.com/mcp

Notes

This project still assumes the upstream CFT app exposes these endpoints:

  • /api/health

  • /api/market

  • /api/deals

  • /api/admin

  • /api/verify

If any of those routes differ on the Vercel app, the MCP server will deploy correctly but those specific tools will fail at runtime.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    • Faire MCP server for ChatGPT, Claude and any AI agent: list orders, update inventory, create reorder and reply to message in your own Faire account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Faire yourself; Cotlet does not store your password and you need no Faire API key. About the official Faire API: Yes, but only part of it: Faire has an External API for brands that covers products, orders and inventory, and a single brand can generate its own access token in the brand portal (checked 11 Oct 2026). We found no API for Messages, Campaigns, your Faire Direct link or your customer list. Setup and FAQ: https://cotlet.dev/integrations/faire-api-mcp-server

      https://cotlet.dev/mcp/v1/faire
    • Tradify MCP server for ChatGPT, Claude and any AI agent: create job, create quote, send invoice, schedule staff and list unpaid invoices in your own Tradify account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Tradify yourself; Cotlet does not store your password and you need no Tradify API key. About the official Tradify API: Tradify publishes no API, developer docs or webhooks (checked 11 Oct 2026). It connects to other tools through built-in integrations: QuickBooks, Xero, Sage, Stripe and Google Calendar. Setup and FAQ: https://cotlet.dev/integrations/tradify-api-mcp-server

      https://cotlet.dev/mcp/v1/tradify
    • Freshsales MCP server for ChatGPT, Claude and any AI agent: list rotten deals, move deal to another stage, mark deal won or lost, create task, add note, list overdue tasks, list contacts and list pipelines in your own Freshsales account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Freshsales yourself; Cotlet does not store your password and you need no Freshsales API key. About the official Freshsales API: Yes, Freshsales has a full REST API, but no official MCP server for CRM data (checked 22 Oct 2026). You copy a personal API key from Profile Settings > API Settings and send it with every call, and Freshsales lists API access as a paid-plan feature. Setup and FAQ: https://cotlet.dev/integrations/freshsales-api-mcp-server

    • Fakturownia (InvoiceOcean) MCP server for ChatGPT, Claude and any AI agent: issue invoice, send invoice, mark invoice paid, add client and list unpaid invoices in your own Fakturownia (InvoiceOcean) account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Fakturownia (InvoiceOcean) yourself; Cotlet does not store your password and you need no Fakturownia (InvoiceOcean) API key. About the official Fakturownia (InvoiceOcean) API: Yes. Fakturownia has a documented REST API that creates invoices, clients and payments, sends invoices by email and sends them to KSeF (checked 14 Oct 2026). Setup and FAQ: https://cotlet.dev/integrations/fakturownia-api-mcp-server

    Related MCP Servers

    • A
      license
      B
      quality
      D
      maintenance
      Enables AI assistants to interact with WhatsApp through the WAHA (WhatsApp HTTP API) platform. Supports chat management, message operations including sending/receiving messages, and marking chats as read.
      52
      11
      ISC
    • A
      license
      B
      quality
      C
      maintenance
      Bridges the WhatsApp HTTP API with AI assistants to enable full control over messaging, chat management, and interactive workflows through 63 specialized tools. It allows users to automate WhatsApp tasks and receive real-time AI feedback directly on their mobile devices.
      63
      8 npm
      6
      MIT
    • A
      license
      B
      quality
      D
      maintenance
      Enables AI agents to control WhatsApp, including sending messages and media, reading chats, managing groups and communities, with QR/pairing auth and session persistence.
      14
      22 npm
      2
      MIT