Summary

Your existing automations become the tool list.

There is no tool authoring here at all: whatever you have already built in Make, set to On-Demand, is what the assistant can call — and it reads the scenarios' own parameter definitions so the descriptions stay in sync with the automation. The one thing to check before adopting it is the notice at the top of the README, which points most users at Make's newer cloud-based server.

What it is

An MCP server that connects to your Make account, finds every scenario configured with On-Demand scheduling, and exposes each one as a tool. It parses and resolves each scenario's input parameters so the assistant gets meaningful descriptions rather than raw field names, and returns scenario output as structured JSON. Make now publishes a cloud-based version and recommends it for most use cases; this repository is labelled the legacy server.

What you get

Every On-Demand scenario in your team surfaced as a callable tool, without writing a tool definition for any of them
Input parameters parsed and resolved into descriptions an assistant can act on
Scenario output returned as structured JSON, so results can be interpreted rather than scraped from text
Bidirectional use: automation logic stays in Make, while the AI side gains the ability to trigger it and read what came back

Requirements

NodeJS and an MCP client. You need a Make API key with the `scenarios:read` and `scenarios:run` scopes, generated in your Make profile, plus `MAKE_ZONE` (the zone your organisation is hosted in, such as `eu2.make.com`) and `MAKE_TEAM`, whose ID is in the URL of the Team page. The package is `@makehq/mcp-server` (0.5.0), run through `npx`. Note the header on the README: a modern cloud-based Make MCP Server now exists and is recommended for most use cases.

Setup effort

One command plus a key — npx -y @makehq/mcp-server, then supply credentials