How to Connect AI Agents to Make Using MCP

A practical field guide from Automation Ace.

Short answer: Make supports MCP in both directions. The Make MCP Server exposes your scenarios as tools, so AI clients such as Claude, ChatGPT, or Cursor can run them with structured inputs; each scenario must be active and set to run on demand, and you connect with an MCP token from your Make profile. The Make MCP Client, and MCP tools in Make AI agents, let your scenarios and agents use tools from other MCP servers. Make has said both are included on all plans at no extra cost; confirm current terms.

Short answer

• Outside AI → Make: Make MCP Server, scenarios as tools.
• Make → outside tools: Make MCP Client and MCP tools in agents.
• Keep it safe: narrow scenarios, clear inputs, least privilege, and logging.

MCP in one minute

The Model Context Protocol is a standard way for AI applications to discover and call tools. Instead of a custom integration for every assistant, a tool provider runs an MCP server, and any MCP client can use it. See API vs CLI vs MCP vs SDK and Zapier MCP for a comparable approach.

Three ways MCP works with Make

Let outside AI agents call Make scenarios

Expose a scenario to an outside AI client (Make MCP Server)

1. Build a scenario that does one job, e.g. "Create HubSpot task".
2. Set it to run on demand and make it active.
3. Define clear scenario inputs (e.g. contact_email, task_title, due_date)
   with descriptions, so the AI knows what to send.
4. Return a clear output (e.g. task_id, status).
5. In your Make profile, open API / MCP access and create an MCP token.
6. Add the Make MCP server to your AI client with that token.
7. Ask the client to list tools, then test with a harmless request.

Check Make's MCP documentation for the exact menu names and supported clients; the setup has evolved.

Let Make agents use outside MCP tools

1. Find the MCP server URL and authentication for the external tool.
2. In the Make AI agent builder, add the MCP server (Make's newer agents provide an option to add MCP tools).
3. Choose which of its tools the agent may use, if the server offers many.
4. Describe in the agent's instructions when to use those tools.
5. Test and review tool calls in run details.

See how to build AI agents in Make.

Design scenarios that make good tools

• One job per scenario: “create task,” not “do CRM stuff.”
• Clear names and descriptions so the AI picks the right tool.
• Typed, described inputs with validation at the start.
• Clear outputs the AI can report back.
• Error handling that returns a useful message; see Make error handling.
• Idempotency so repeated calls do not duplicate records.

Security and permissions

• Expose only the scenarios an AI client needs; prefer read-only tools first.
• Treat the MCP token like a password; rotate it if exposed.
• Use connections with the least access the job needs.
• At Waves '26, Make described checking agents that connect over MCP against their owner's role at the moment they act; see controlling AI agent access and permissions.
• Add human approval for irreversible actions.

Testing and monitoring

Test each tool with harmless inputs, then with edge cases. Watch execution history for MCP-triggered runs and track volume and failures; see monitoring AI agents in Make.

Use cases

• A team member asks Claude to “create a follow-up task for this client,” which runs a Make scenario.
• A coding assistant triggers a deployment-notification scenario.
• A Make support agent uses an external knowledge tool over MCP.
• Several agents share the same scenario tools; see coordinating multiple agents.

Frequently asked questions

How do I connect an AI agent to Make using MCP?

To let an outside AI client use Make, build focused scenarios set to active and on demand with clear inputs and outputs, create an MCP token in your Make profile’s API/MCP access section, and add the Make MCP server to the client. To let Make agents use outside tools, add the external MCP server in the Make AI agent builder.

What is the Make MCP Server?

The Make MCP Server exposes your Make scenarios as tools that MCP-compatible AI clients such as Claude, ChatGPT, or Cursor can call with structured inputs.

What is the Make MCP Client?

The Make MCP Client lets Make scenarios or Make AI agents connect to other MCP servers so your automations can use tools hosted outside Make.

Does Make MCP cost extra?

Make has said the Make MCP Server and Make MCP Client are included in all Make plans at no extra cost; scenario runs still use credits. Confirm current terms with Make.

How do I keep Make MCP secure?

Expose only the scenarios a client needs, start with read-only tools, protect and rotate the MCP token, use least-privilege connections, validate inputs, and require approval for irreversible actions.