Make MCP Server

The Make MCP Server connects your AI assistant to your account on Make, the no-code automation platform. MCP, the Model Context Protocol, is an open standard that lets assistants such as Claude, ChatGPT or Cursor reach external services. Make runs the server itself in the cloud, so there is nothing to install. The offering comes from Make, based in Prague, which used to be known as Integromat and is now part of Celonis. In Make, automated workflows are called scenarios. A scenario might read a form, create a record in your CRM and post a message to Slack.

What is the Make MCP Server good for?

The server is built on a simple idea: every scenario you have prepared in Make becomes a tool for the AI. You build a workflow once, cleanly and tested, in Make. After that, the assistant can start it on request in the chat and work with the result. This is more reliable than letting the AI piece together dozens of API calls on its own. A sales team could, for example, say: create this lead and send the welcome email. Behind the scenes, exactly the scenario the department approved will run. With more than 3,000 connected apps, the assistant turns into a front end for a company's entire automation setup.

Scenarios as tools

For a scenario to show up as a tool, it has to be active and use on-demand scheduling, meaning it only runs when requested. Scenario inputs and outputs define which data the assistant sends and which it gets back. Make recommends describing each scenario in detail. The AI picks the right tool based on that description. According to the docs, a run may take 25 seconds over OAuth and 40 seconds with a token. If it takes longer, the scenario keeps running for up to 40 minutes. The assistant then receives an execution ID and can fetch the result later.

Management tools

Besides running scenarios, the server offers tools for managing your account. With them, the AI can view and change scenarios, maintain connections, webhooks and data stores, and manage teams and organizations. According to Make, it can even invite members to an organization. These tools are only available on paid plans. Which tools the assistant gets depends on the scopes. Scopes are permissions that you pick one by one when you connect.

Connecting via OAuth or token

The recommended route is OAuth at mcp.make.com. OAuth is a sign-in method where you grant access in your browser without handing over a key. During sign-in, you also pick the organization the assistant may access. According to Make, only OAuth lets you restrict access to one specific organization. Alternatively, you can connect with an MCP token through the address of your Make zone, such as eu2.make.com. This route allows longer timeouts. Make documents separate guides for Claude, Claude Code, ChatGPT, Cursor, Windsurf, Gemini CLI, Kiro, Warp and Mistral AI. In Claude, Make is listed as a ready-made connector in the directory.

Free plan and paid plans

According to Make, running scenarios works on every plan, including the free one. The management tools require a paid plan. Make bills in credits, and most actions in a scenario use one credit. Every run triggered by the assistant therefore counts like a normal scenario run. You can find the current plans, from Free through Core, Pro and Teams to Enterprise, on Make's pricing page.

Who is the Make MCP Server for?

The server pays off for anyone who already automates with Make and now wants to trigger those workflows from a chat. That covers operations, marketing and sales teams, as well as agencies that look after scenarios for clients. Its edge over letting the AI drive apps freely is control. The assistant only runs what you have built and tested beforehand. If you have no scenarios yet, the server alone will not get you far. Compared with the Zapier MCP Server or the n8n MCP Server, Make is the right choice mainly when your team already works in Make's visual scenario editor. Keep in mind that the run scope opens up every active on-demand scenario in the organization. According to Make, you can separate access more finely from the Teams plan up, using dedicated accounts per team.

A Make account, free or on a paid plan; the management tools need a paid plan. At least one active scenario with on-demand scheduling so that tools show up. Plus a client that supports remote MCP over Streamable HTTP, such as Claude, ChatGPT, Claude Code or Cursor.

With the Make MCP Server, your AI assistant runs Make scenarios as tools and, if you like, manages your Make account. Make hosts the server, so you install nothing. This guide shows how to prepare scenarios and connect the server in common clients.

Prepare your scenarios

A scenario only shows up as a tool if it is active and uses on-demand scheduling. Use the scenario inputs to define what data the assistant should send, and the outputs to define what it gets back. Write a clear description for the scenario. The AI uses it to decide when to call the tool.

Claude

Make is listed in Claude as a ready-made connector. Open Customize in the sidebar, then Connectors and Discover. Search for Make and click the plus sign. Pick the organization and scopes in the consent dialog and click Allow.

Step 1: Claude Code

Run claude mcp add --transport http make https://mcp.make.com/claude in your terminal. Then start Claude Code with claude and call /mcp. Make's consent dialog opens in your browser, where you pick the organization and scopes. Claude Code then shows the server as connected.

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

Cursor

Open the MCP tools section in the Cursor settings and add a custom server. In mcp.json, add a make entry with the url https://mcp.make.com/cursor. Then click Needs login next to Make and confirm the sign-in in your browser.

Step 2: Other clients and MCP tokens

Connect other clients with remote MCP support through https://mcp.make.com/v2. Make has dedicated guides for ChatGPT, Windsurf, Gemini CLI, Kiro, Warp and Mistral AI. If OAuth gives you trouble, create a token with the mcp:use scope under Profile and API access. Then use your zone's address, such as https://eu2.make.com/mcp/v2, with the token in the Authorization header.

Timeouts and long runs

Over OAuth, a scenario run may take 25 seconds; with a token, 40 seconds. If it takes longer, the scenario keeps running for up to 40 minutes. The assistant receives an execution ID and can fetch the result later, provided the scope for reading scenarios is granted.

Work safely

The run scope opens up every active on-demand scenario in the organization. So only set scenarios to on-demand if the AI should really be able to start them. Grant management scopes only when you need them. A token in the URL is as secret as a password and does not belong in shared configs.

OAuth with your Make account, with organization and scopes picked in the consent dialog; only this route can restrict access to one organization. Alternatively, an MCP token from Profile > API access with the mcp:use scope, passed as a bearer header or in the URL.

Depending on scope: view and run active on-demand scenarios and fetch their results. On paid plans, also view and change scenarios, connections, webhooks and data stores, and manage teams and organizations, including invitations.

Data flows through the Make cloud and the apps connected in your scenarios. Make's documentation gives no retention details specific to the MCP server. Content in the chat also goes to the respective AI provider.

Medium. Scenarios can write to many connected apps, and the management tools change scenarios and connections. A crafted text with hidden instructions can trigger unwanted runs. Grant only the scopes you need, keep scenarios narrow and treat a token in the URL like a password.