Make's MCP server gives any AI — Claude, ChatGPT, Cursor — direct access to the scenarios already running inside your Make account. No new platform, no separate integration layer. The AI calls your scenarios the same way a human clicking "Run" would, and the results come back structured and ready to use.
It's included in every Make plan, free through Enterprise. The cost model is simple: you pay only the Make credits consumed by the scenarios the AI actually runs. The MCP access itself costs nothing extra.
What Make MCP does (and how it differs from Zapier MCP)
The MCP (Model Context Protocol) standard lets AI clients — Claude, ChatGPT, Cursor, and others — call external tools through a consistent interface. Make and Zapier both have MCP servers, but they work differently, and which one fits depends on how you build.
Zapier MCP is agentic. Connect an AI to Zapier's MCP server and it gets free-range access to 9,000+ apps on demand. No Zap built in advance — the AI decides what to do in real time based on your prompt.
Make MCP is structured. The AI can only call scenarios you've already built inside Make. You expose specific scenarios as MCP tools; the AI calls them with the inputs you defined. Nothing the AI can trigger falls outside the guardrails of a scenario you designed and tested.
That tradeoff matters in practice:
| | Make MCP | Zapier MCP | |---|---|---| | What the AI can do | Call pre-built Make scenarios | Any of 9,000+ apps on demand | | Setup time | Higher — scenarios must exist first | Fast — connect once, prompt freely | | Control | High — AI is bounded by your scenario logic | Lower — AI decides what actions to take | | Cost at high volume | Often cheaper — credit-based, no per-step uplift | Can exceed $300/mo at 100k operations | | Best for | Complex logic, multi-step transforms, cost-conscious | Speed, breadth, non-technical clients |
For operators selling services, the choice usually comes down to client type. Technical clients with complex workflows (multi-step data transforms, error-handling branches, conditional routing) benefit from Make MCP's structure. Non-technical clients who want AI that "just handles things" without pre-built scenarios lean toward Zapier.
Many operators use both: Zapier MCP for exploratory and fast-deployment client work, Make MCP for the production scenarios they've refined and want AI to trigger reliably.
What's included in the Make MCP server
Make's MCP server exposes two categories of tools to connected AIs:
Scenario execution tools (available on all plans, including free): the AI can run your scenarios. You expose specific scenarios as callable tools with named inputs and outputs, and the AI triggers them the same way a webhook would.
Management tools (paid plans only): the AI can view and modify your Make account — list scenarios, check execution history, update scenario settings. Useful for agent systems that monitor and manage automations, not just run them.
The server is cloud-hosted by Make (no local setup, no separate server to maintain) and uses Stateless Streamable HTTP as its default transport, which means it works reliably with any MCP client without persistent connection requirements.
How to connect your AI client
Setup takes under 15 minutes:
- In Make, go to Settings → MCP Server and generate your server credentials
- Choose which scenarios to expose — you pick exactly which ones become AI-callable tools
- Copy the MCP server URL and authentication token
- Paste into your AI client's MCP settings (Claude Desktop, Claude Code, ChatGPT, Cursor, or any MCP-compatible client)
The scenario selection step is important: you define the inputs each scenario accepts and what it returns, and those become the tool's API. A well-named scenario with clear inputs ("create_client_record: client_name, email, project_type → returns record ID") is much easier for an AI to call correctly than one named "Scenario 47."
What it costs
Make MCP access is free on every plan. You pay only for the Make credits consumed by scenarios triggered through MCP — the same credit cost as triggering the scenario any other way.
Credit consumption follows Make's standard model: each module operation in a scenario costs 1 credit. A 5-module scenario costs 5 credits per run whether a human triggers it or an AI does via MCP.
| Make plan | Annual price | Credits/mo | MCP scenario calls at 5 modules | |---|---|---|---| | Free | $0 | 1,000 | ~200 | | Core | $9/mo | 10,000 | ~2,000 | | Pro | $16/mo | 10,000 + flex | ~2,000+ | | Teams | $29/mo | 10,000 + flex | ~2,000+ |
At Core pricing ($9/mo, annual), a production client deployment running 50 AI-triggered scenario calls per day at 5 modules each burns 7,500 credits — well within the 10,000/mo plan, with no MCP surcharge. The economics look very different than Zapier at the same volume, where 50 daily calls × 2 tasks each = 3,000 tasks/month, requiring at least the Professional plan ($19.99/mo).
The service you sell with it
Make MCP's value for operators is building AI-accessible automation systems for clients — and charging for the design, not just the execution.
The offer: "AI-activated operations." You build a library of Make scenarios that cover the client's recurring processes (onboarding, billing, reporting, routing), then connect their AI assistant to trigger any scenario on command. The client stops maintaining manual SOPs for routine work.
The flagship build: an AI assistant that handles client onboarding end-to-end. When a new client signs, the AI triggers three Make scenarios in sequence: create the CRM record, send the welcome sequence, provision access to the client portal. No human in the loop for the mechanics.
What to charge:
- Design and build: $2,500-$5,000 — scenario architecture, MCP configuration, AI prompt library, testing and documentation
- Retainer: $400-$800/mo — monitor scenario health, add new AI-callable scenarios as the client's operations evolve, track credit usage
The build cost is higher than a Zapier MCP deployment because the upfront scenario design is part of what you're selling — a custom, AI-accessible operations layer, not just a generic connection. That's also what earns the premium retainer. Clients who understand what they're buying (an AI that reliably executes their specific processes, not a general-purpose agent) renew at higher rates because the dependency is real.
For the retainer math on automation services generally, see how to start an AI automation agency — the packaging logic is identical.
FAQ
Is Make MCP free?
Make MCP server access is included in every Make plan, including the Free plan. You pay only for the Make credits consumed when scenarios actually run. There's no MCP surcharge, no add-on fee, and no per-call cost on top of your plan.
How is Make MCP different from Zapier MCP?
Make MCP requires pre-built scenarios that the AI calls on demand. Zapier MCP lets the AI take ad hoc actions across any app without a pre-built flow. Make gives you more control and lower running cost at high volume; Zapier gives you speed and breadth with less upfront setup. See the comparison table above.
What AI clients work with Make MCP?
Any MCP-compatible client: Claude Desktop, Claude Code, ChatGPT (via Developer Mode), Cursor, GitHub Copilot, Windsurf, and others. Make's server uses Streamable HTTP transport, which is supported by all major MCP clients.
Do I need coding skills to set up Make MCP?
No. Setting up the MCP server itself (generating credentials, copying the URL, pasting into an AI client) requires no code. Building the Make scenarios that the AI will call requires Make's visual builder — no code, but some automation logic experience helps. The more complex your scenario, the more your design decisions matter.
Can I use Make MCP for client work?
Yes — this is the high-margin use case. Build scenarios in the client's Make account, expose them as AI tools via MCP, connect their AI assistant (Claude or ChatGPT), and charge for the system design and maintenance. Client Make plans start at $9/mo — invisible relative to your service fee. Stack Make's 35% affiliate commission on the referral and the economics improve further.