The Trengo MCP server connects your Trengo inbox to an AI assistant. It works with Claude, ChatGPT, Cursor, VS Code and other tools that support the Model Context Protocol.

The assistant can then work in your inbox: read and answer conversations, update contacts, build AI agents and journeys, and pull reports. It does this work as you. It gets the same access that your Trengo user has, and no more.

Trengo hosts the server. You install nothing. You create no API key.

https://mcp.trengo.com/mcp

What you can ask it to do

There are 88 tools. A tool that needs a label, team, agent, channel or custom field takes its name. You do not need to know the ID.

These are examples of what you can ask:

• "Show the conversations that wait on us, longest first."
• "Read conversation 48213. Write a reply about our refund policy."
• "Add an internal note to the last WhatsApp conversation with Anna. Ask billing to check invoice 2024-119."
• "How many conversations did the Sales team close yesterday?"
• "Find the contact for +31 6 12345678. Add the contact to the VIP group."
• "Create an AI agent for our webshop questions and test it with five customer questions."

To see the full list, ask the assistant. It reads all the Trengo tools when it connects.

Before you start

• You have a Trengo account. The assistant gets the access of that account.
• Your assistant supports remote MCP servers with OAuth. The list below gives the steps for the most common ones.
• You are signed in to Trengo in your browser. The consent screen needs this.

An administrator does not have to create an app first. Your assistant registers itself when you connect.

Connect your assistant

Claude (web and desktop)

1. Open Settings > Connectors.
2. Click Add custom connector.
3. Type the name Trengo. Paste the address https://mcp.trengo.com/mcp.
4. Click Connect. Trengo shows a consent screen. Read it, then click Allow.
5. Open a chat. Open the tools menu. Switch Trengo on.

Claude Code

claude mcp add --transport http trengo https://mcp.trengo.com/mcp

Type /mcp in Claude Code. Select Authenticate. Your browser shows the same consent screen.

Cursor

Open Cursor Settings > MCP & Integrations > New MCP server. You can also edit the file ~/.cursor/mcp.json:

{
  "mcpServers": {
    "trengo": {
      "url": "https://mcp.trengo.com/mcp"
    }
  }
}

Cursor shows the status Needs login. Click it and complete the consent screen. The tool list then becomes green.

To use the connector in one project only, put the same file at .cursor/mcp.json in that project.

ChatGPT

Connectors that call MCP tools are part of developer mode. Your plan and your workspace administrator control access to it.

1. Open Settings > Connectors > Advanced. Switch Developer mode on.
2. Go back to Connectors. Click Create.
3. Type the name Trengo. Paste the address https://mcp.trengo.com/mcp. Select OAuth.
4. Click Create. Complete the Trengo consent screen.
5. Open a chat. Open the + menu. Switch the Trengo connector on.

VS Code (GitHub Copilot)

Add the file .vscode/mcp.json to your workspace:

{
  "servers": {
    "trengo": {
      "type": "http",
      "url": "https://mcp.trengo.com/mcp"
    }
  }
}

Start the server from the CodeLens above the entry. Complete the consent screen in your browser.

Other MCP clients

Give the client the address https://mcp.trengo.com/mcp. The client finds everything else.

Disconnect an assistant

To disconnect the assistant, remove the connector in the client.

See also

• REST API introduction
• Model Context Protocol specification