Make.com + Claude Desktop: MCP Setup Guide for Agentic Automation

Contains affiliate links. Disclosure.

Most "AI automation" demos stop at chat output. The real value starts when your assistant can trigger controlled business actions. This guide shows a practical Make.com plus Claude Desktop setup using MCP (Model Context Protocol) so Claude can call Make workflows safely.

If you need a Make account first: Start here. New to Claude and want to test this without committing? You can try Claude free for a week.

What MCP does in this setup

MCP (Model Context Protocol) is an open standard for connecting AI applications to external tools and systems. In this flow:

• Claude acts as the MCP client.
• Make exposes callable tools/workflows via its MCP capabilities.
• You control what actions are available and what data can be accessed.

Prerequisites

• Claude Desktop with MCP support. MCP connectors work on the free tier, but for regular automation use the message limits make Pro or higher more practical.
• Node.js (for npx mcp-remote)
• Make workspace with at least one test scenario
• MCP token from Make profile/settings

Architecture overview

1. User asks Claude to perform a business action.
2. Claude calls an allowed tool via MCP.
3. Make runs the mapped scenario.
4. Scenario output returns to Claude as structured result.

Step 1: Get your Make MCP token and zone

1. Log into Make.
2. Go to profile/settings and find MCP or Developer section.
3. Generate an MCP token.
4. Note your zone (e.g., eu2.make.com, us1.make.com) from the URL or docs.

Step 2: Prepare one Make scenario as a tool target

Start simple:

• Trigger: webhook or a clearly parameterized entry point.
• Action: one safe destination action (for example, add row / send alert).
• Output: structured confirmation (status, id, timestamp).

This keeps testing deterministic.

Step 3: Connect Claude to Make MCP

Make exposes a remote MCP endpoint. The most reliable way to wire it into Claude Desktop is the mcp-remote bridge in claude_desktop_config.json, pointing at your zone's MCP URL with your token. Make's token-based SSE endpoint follows the pattern https://<MAKE_ZONE>/mcp/u/<MCP_TOKEN>/sse:

{
  "mcpServers": {
    "make": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://eu2.make.com/mcp/u/<your-mcp-token>/sse"]
    }
  }
}

Substitute your actual zone (for example eu2.make.com or us1.make.com) and MCP token. Make also supports OAuth sign-in via https://mcp.make.com if you would rather not paste a token. After saving, fully quit and relaunch Claude Desktop. MCP servers load only at startup.

Config file location:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Make's exact endpoint format and any native connector UI can change, so confirm the current URL pattern against Make's MCP documentation (developers.make.com/mcp-server) before pasting it in.

Step 4: Define strict tool boundaries

Do not expose broad "do anything" operations first.

Start with:

• One scenario.
• Narrow parameter schema.
• Explicit allowed fields.
• Logging enabled.

Example safe tool contract:

{
  "lead_name": "string",
  "company": "string",
  "priority": "low|medium|high",
  "action": "create_note"
}

Step 5: Test from Claude Desktop

Run three tests in sequence:

1. Happy path: valid inputs.
2. Validation fail: missing required field.
3. Policy fail: disallowed action.

You want predictable errors and clear confirmations, not silent fallback behavior.

Step 6: Add one production guardrail

Before wider rollout, add at least one of:

• Human approval step for high-risk actions.
• Confidence threshold for AI-generated routing.
• Region-based policy checks before external communication.

Common setup issues

• Connector added but tool not visible → verify scenario publication/permissions.
• Claude call times out → simplify scenario and reduce upstream latency.
• Output parsing errors → enforce stable JSON schema.
• Wrong account context → verify environment/workspace mapping.

Compliance considerations

When Claude can trigger business workflows, compliance risk increases if scope is too broad.

• Restrict tool scope by region and business unit.
• Avoid exposing personal data fields unless required.
• Keep audit logs for prompts, tool calls, and final actions.
• Apply consent and communication rules by target market.

Next step

Build one narrow tool first, validate with real logs, then expand. Fast, controlled rollout beats broad unstable rollout.

Create your first Make MCP-ready scenario.

Production rollout checklist

• Tool schema documented.
• Error handling tested.
• Access policies defined.
• Region/compliance checks implemented.
• Monitoring alerts enabled.

Conclusion

A reliable claude desktop mcp setup with Make is about control, not just connectivity. Start with one low-risk workflow, enforce strict schemas, and expand only after logs prove stability.

Next step

After your first MCP tool works, add one read-only reporting tool before adding more write actions. Start building with Make MCP →

Frequently Asked Questions

Yes. With Make's MCP server and proper config, Claude can run Make scenarios as tools. Add the MCP server to claude_desktop_config.json and restart Claude.

Basic setup can be done with low-code/no-code patterns, but technical knowledge helps for schema, auth, and error handling.

Expose one narrow, low-risk tool first with strict input validation and full logs before enabling broader actions.

Scenario run tools are available on all plans. Management tools (view/modify teams, scenarios) may require paid plans. Check Make MCP docs.