Last week, Make.com added MCP (Model Context Protocol) support to AI Agents. This means you can now expose Make.com scenarios as tools callable by Claude Desktop, Cursor, ChatGPT, and any other MCP-compliant LLM client.
This unlocks a powerful pattern: your Make.com scenarios become tools that LLMs can invoke autonomously. Instead of going to Make.com to run a workflow, your LLM client runs it on your behalf when needed.
Here's a step-by-step tutorial showing how to expose a Make.com scenario as an MCP tool. I'll use a simple example — searching your Pipedrive CRM for contacts — but the pattern works for any scenario you've built.
Prerequisites
- Make.com Pro or Teams plan (AI Agents feature required)
- An MCP-compliant client. I'll use Claude Desktop in this tutorial. Cursor, Continue.dev, and others also work.
- At least one working Make.com scenario you want to expose
Step 1: Identify the scenario to expose
For this tutorial, I'm using a scenario called "Search Pipedrive contacts" that takes a search query (e.g., "Acme Corp") and returns matching contacts as JSON. Any scenario with a clear input/output structure works.
Good candidates for MCP tools:
- Read-only queries (CRM lookups, knowledge base searches, inventory checks)
- Status checks (order tracking, deployment status, project state)
- Lightweight writes (logging notes, creating draft items)
Avoid exposing scenarios that:
- Take destructive actions (deletions, billing changes, sending emails to customers)
- Have high operations cost (will burn through your quota fast)
- Require complex authorization decisions
Step 2: Configure the scenario as MCP tool
In your Make.com scenario settings, look for the new "MCP" tab (added last week). You'll see:
- Tool name: short identifier like
search_pipedrive_contacts(no spaces, snake_case) - Description: what the tool does, written for the LLM. E.g., "Searches Pipedrive CRM for contacts matching a name or company. Returns up to 10 matches with name, email, company, and last activity."
- Input schema: JSON Schema describing parameters. For our example:
{"query": "string, required"} - Output format: JSON Schema for the response
Important: the description matters a lot. The LLM uses it to decide whether to call your tool. Be specific about what it does and when it should be used.
Step 3: Generate the MCP server URL
After saving the MCP configuration, Make.com generates a unique URL like:
https://mcp.make.com/your-org-id/tools/search_pipedrive_contacts
This URL is your MCP endpoint. Make.com also generates an authentication token — keep it secure.
Step 4: Add the tool to Claude Desktop
Open Claude Desktop config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Add your Make.com MCP server to the mcpServers section:
{
"mcpServers": {
"make-pipedrive": {
"command": "npx",
"args": ["mcp-remote", "https://mcp.make.com/your-org-id/tools/search_pipedrive_contacts"],
"env": {
"MCP_AUTH_TOKEN": "your-token-here"
}
}
}
}Restart Claude Desktop. The tool should now appear in the available tools list.
Step 5: Test it
In Claude Desktop, ask: "Can you search my Pipedrive for contacts at Acme Corp?"
Claude will recognize the tool, call your Make.com scenario, and return the results. The first call may take 2-3 seconds as Make.com warms up the scenario; subsequent calls are faster.
Common errors and fixes
"Tool not found" — Restart Claude Desktop after editing the config. Check the config JSON is valid (use jsonlint.com).
"401 Unauthorized" — Wrong MCP_AUTH_TOKEN. Regenerate in Make.com and update.
"Tool calls fail intermittently" — Likely operations quota issue. Check your Make.com usage. Each tool call costs 5+ operations.
"Output doesn't make sense to Claude" — Improve your tool description and output schema. The LLM needs structured, predictable data.
Going further
This basic pattern can be extended significantly:
- Expose multiple scenarios as a "toolkit" (CRM lookup, deal creation, note logging)
- Add input validation to prevent expensive or dangerous calls
- Return rich data with embedded references (contact ID, Pipedrive URL) so Claude can suggest follow-up actions
- Combine with Claude Projects to give your team shared access
If you're building this for production use across a team, our Custom Make.com Integration service handles MCP-based deployments end-to-end.
Why this matters
MCP changes how we think about Make.com scenarios. Previously, scenarios were workflows you ran by trigger. Now, they can be tools your AI assistant runs autonomously, on your behalf, when needed.
This is a fundamental shift toward agentic automation. Start small (one read-only scenario as a tool), see how it changes your workflow, then expand.