A few weeks ago on hosting.com's Agency Success Office Hours, Ben B. asked a question that really has got me thinking "With the epidemic of MCP connectors to Claude and other AI tools, do you think this is the year to build out workflows to replace Zapier?"
Nathan Ingram gave a good answer: “Zapier still has its place, MCP routines feel early, don't rip out what's working.”
He's right on the practicalities. As some would say, “if it ain’t broke don’t fix it!”.
But I want to add a layer to that conversation, because the way I see this playing out isn't really "MCP vs. Zapier". Instead, I believe it's all about which layer of your workflow each tool belongs in and what you're paying for.
As I have been spending most of my evenings seeing what opportunities I can unlock with tools like Claude Code, Cursor, Lovable, (and anything else I can get my hands on), I have been truly able to break through to a new understanding of what the future of automation and the agentic web really looks like.
Here's my take. Zapier's days aren't numbered. They're just getting more specific.
What Zapier is doing right for marketers
It's undeniable that Zapier does a lot of things well.
For most in-house marketing teams, Zaps are running the following work:
Lead capture and CRM updates: Form submissions, ad leads, and chat captures landing in HubSpot or Salesforce with the right tags and source attribution.
Content distribution: New blog post triggers social posts across LinkedIn, X, and Facebook.
Email triggers: Post-purchase review requests, cart abandonment sequences, win-back flows.
Reporting plumbing: Pulling metrics from Google Analytics, ad platforms, and social into a shared spreadsheet or dashboard.
Threshold alerts: CPA exceeds a number, conversion drops below a number, Slack gets pinged.
Data syncing: Keeping spreadsheets, CRMs, and email tools aligned so nobody's working from stale records.
Client onboarding and ops: Welcome sequences, task creation in project management tools, dashboard provisioning.
This is a lot of work, and Zapier does it affordably and reliably.
Where MCP is changing the game
Before we begin diving deeper, it’s important to understand what MCPs are (or Model Context Protocol). Simply put, it’s the standard that lets AI tools like Claude talk directly to your other applications.
I personally use Claude Code as my tool of choice, as it is where I can make the most impact cost effectively, but this can be any tool of your choosing. I have also been using Ahrefs and Google products very regularly in regard to MCPs – however, the possibilities are expanding very quickly. You can connect to your CMS, your CRM, and so much more.
And the reason it is very exciting to a lot of people, is that instead of building rigid step-by-step automations, you give an AI model the ability to reach into a tool and decide what to do based on what it finds.
That distinction is everything. With Zapier you define every step, every condition, every fallback. Additionally, MCP-connected AI is reasoning-based. The model looks at the situation and chooses the next move using the data, prompt, skills, and anything else you have supplied it with. It works like your newest team member or someone who understands you, your goals, and your business.
Here's what that looks like in practice
Below are four workflows you can implement today. Each one represents a task that either couldn't be automated reliably or was being done manually because the variables never quite fit a fixed rule. These are the jobs where judgment matters.
Landing page generation at speed:
We use the Figma MCP to streamline landing page creation, and the key is that we don't just point an AI at Figma and hope for the best. We give it rules and context to work within. This includes our brand fonts and how they're applied, our color palette and where each color belongs, hover states and interaction patterns, imagery guidelines, tone of voice for body copy and CTAs, spacing and layout principles.
The AI builds within those constraints rather than guessing at them. That means every page comes out on-brand by default, which lets us test and optimize at a velocity that wasn't realistic before. The bottleneck used to be production designers and developers building each variant from scratch.
Now the bottleneck is deciding what to test next, which is a much better problem to have.
Cross-source data analysis:
Our Ahrefs MCP and Google MCP work together. I can ask a question that previously needed a spreadsheet jockey and an afternoon something like "Which competitors gained ground on our top 50 keywords this quarter, and what content did they publish to do it?" - and get an answer with reasoning attached. Not just data dumped into a row. Actual analysis. I can also send these same agents to any website – not just my own – allowing us to truly dig into competitors and remove a huge part of manual analysis that could take humans days.
Content creation feeding into team collaboration:
Our agents use data from analysis (such as what was referenced above) and then draft content, that lands automatically in our Google Drive. The team can then elevate it together adding the experience, expertise, authoritativeness, and trust signals that E-E-A-T demands. The whole process lives in one workspace. We're not exporting, re-importing, copy-pasting between five tools.
Customer service triage with Orbi:
Orbi is our customer support chatbot, and it's a good example of where AI judgment pays off. Customers asking simple questions (account changes, billing, status checks) get resolved in 30 seconds. Customers who need a human get routed straight to one. Orbi doesn't try to handle what it shouldn't. That's the design. And that triage decision is exactly the kind of judgment Zapier could never have made.
Webinar attendee intelligence:
I wanted to include this as we are referencing a webinar question! And, in this use case, instead of dumping every webinar attendee into the same nurture sequence, you can use MCP to pull attendance data and score engagement (who showed up, how long they stayed, whether they asked questions, how they responded to polls).
The AI then makes a call: this person belongs in our typical nurture segment, or that person should be tagged for a sales rep to contact within 24 hours.
The triage isn't a rigid rule, but instead, a decision made every time, based on the actual behavior of each attendee. And the best part is that it can change. It benchmarks each webinar dependent on the content and the learnings from the data.
So why isn't this a full replacement story?
A year ago, running everything through an AI agent looked expensive next to a Zap that costs fractions of a cent. That's still true if you're spinning up agent calls for every routine CRM update at scale.
However, if you're already on a Claude Pro subscription at $20 a month, the marginal cost of an MCP workflow for moderate use is essentially zero.
Zapier Pro is also $20 a month and gives you 750 tasks. For technical users who can wire up MCP themselves, the economics have basically converged.
So, it depends on who you are and what you’d like to achieve. If you're a non-technical in-house marketer, Zapier's no-code ease still justifies its cost premium for deterministic work. You don't need to learn anything new to keep your lead capture flowing. That's real value.
If you're technical, or you've got someone technical on the team, MCP earns its cost the moment the work moves above plumbing. Anything that requires:
Looking at multiple data sources and synthesizing them
Making a judgment call instead of following a fixed rule
Generating content or commentary, not just routing data
Adapting when the input changes shape
That's where the spend is justified, because the output is something you couldn't have produced cheaply any other way.
The framework I keep coming back to - use the cheapest tool that does the job well.
Zapier for plumbing on non-technical teams. MCP for intelligence, and increasingly for everyone (sooner than you'd think).
So what should you do on Monday morning?
If you're sitting in an in-house marketing seat, here's the honest recommendation.
Keep your existing Zaps. The lead capture, the social posting, the threshold alerts - they work, they're cheap, and rebuilding them in MCP would burn budget for no gain unless you are adding them to an agentic flow (which you could do after anyways feeding the data into your agentic flow).
But for any new workflow you're about to build, MCP should be the first thing you consider.
The ecosystem has matured faster than the "feels early" caveat we referenced earlier in this article.
The weekly reporting where you wish someone could write the commentary.
The competitive analysis you do quarterly because doing it monthly is too painful.
The lead scoring you wish was smarter than "downloaded a PDF = warm."
These are MCP solutions to start using now.
The mistake I see most teams making isn't picking the wrong tool. It's trying to make one tool do both jobs - forcing Zapier to handle work that needs judgment, or spinning up agents to do work a simple connector could handle for pennies.
Be honest about which layer you're working in, and the choice gets easy.
So, focus on Zapier for the plumbing and MCP for the thinking. And if you've been waiting for a sign to start exploring what your stack could actually do with a reasoning model wired into it, this is it.
