Next Gen Zaps mainly transforms how to create and maintain a Zap.
Since the documentation updated the 18 September 2026, Zapier explains that a client compatible with the Model Context Protocol can build and deploy automation from a natural language request. The system selects connections, assembles steps and publishes workflow in Zapier. The functionality remains in early access and is not presented as a complete replacement of the classic vendor.
Analysis of official documentation, with no testing on a paid account.
In this guide
What changes with Next Gen Zaps
So far, creating Zapier automation usually meant opening the vendor, choosing a trigger, connecting applications, setting actions and testing each step. Next Gen Zaps adds another front door: an agent AI can prepare workflow from an MCP compatible environment.
In the official documentation, Zapier indicates that an MCP client can select the right application connections, cable the different steps and deploy workflow. In other words, the conversation with the agent becomes a design layer above Zapier, which remains the engine of execution.
This is more important than just a generation of text. The promise is not only "AI suggests automation": the agent can participate in building an executable workflow. For a small business, an agency or an independent who can explain its process but has poor control of automation publishers, the entry barrier can therefore decrease.
However, a simple principle must be retained: faster generated workflow is not necessarily a better designed workflow. Logical errors, loops, bad field matches or actions triggered at the wrong time are still possible. The gain lies first in the speed of setting up and the ability to iterate in natural language.
How creation works via MCP
MCP means Model Context Protocol. The protocol allows a compatible assistant or agent to use external tools using standardized interfaces. In the case of Next Gen Zaps, the agent does not just explain how to create a Zap: he can use Zapier's capabilities to prepare the workflow.
The scenario then becomes close to this: you describe a need: for example, to recover a new prospect, enrich some data, create a line in a CRM and send a notification: then the agent proposes or assembles the corresponding steps. You can then inspect the result in Zapier.
This approach can become interesting for high-intent requests like "automate my leads", "connect my form to my CRM" or "create automation without learning Zapier". It also makes Zapier more comparable to the new automation tools focused on AI agents, while competition has so far focused on the number of integrations, visual control and cost per execution.
Zapier states that Next Gen Zaps is available as early access on request. For a team, this means that you should not immediately build a critical process around this interface assuming it is already stabilized like the rest of the platform.
Workflow Manager, versions and maintenance
The novelty does not stop at the initial generation. Zapier also documents a Workflow Manager which allows to visualize a Next Gen Zap as a diagram. Three modes of work are announced: describe a natural language modification to the assistant, edit the TypeScript or intervene directly on the visual canvas.
This is probably one of the most useful points for professional users. Serious automation must remain readable by a human. Moving from conversation to diagram and then, if necessary, to code limits the risk of creating a "black box" that cannot be maintained.
Zapier also adds a logic of drafts and versions. A workflow can be modified without immediately publishing the changes; each publication creates a new version. For a company, this allows you to check the changes before you touch a production process and return to history more easily.
Another strong signal: Zapier documents a mode Agentic Management. The agent can monitor the execution of a Next Gen Zap, diagnose certain breakdowns and prepare a correction. Zapier indicates that the level of autonomy can be controlled. This is an important evolution: the AI is no longer only used to create automation, but also to help maintain it.
Access and impact on price: what to check
The official page of Next Gen Zaps viewed the 21 September 2026 indicates an availability in early access on the maps Professional, Team and Enterprise, with a prior access request. The Free plan is listed as ineligible. For Team, the request must come from an owner or administrator.
A distinction must be made between the price of the plan and the actual cost of implementation. Zapier charges his environment around the tasks consumed. On its current tariff page, the vendor explains that workflows, steps AI, connector calls, MCP and SDK share the same task pool, with rates that may vary depending on the complexity and model used.
For the AI by Zapier steps, the pricing model in effect since the 15 June 2026 applies multipliers according to the model level: Standard, Advanced or Premium. The number of tasks consumed can therefore increase rapidly when a scenario adds tool calls or uses more expensive models.
Before choosing Zapier for a recurring process, do not compare only the price displayed per month. Estimate the number of executions, the number of actions per execution and the place of the AI in each run. This is particularly important in the face of Make and n8n, whose economic models and levels of control are different.
For whom this novelty is useful?
Independents and small businesses who know how to describe their process
The main benefit is the reduction of time spent understanding the vendor. If you know how to explain precisely "when X arrives, I want to check Y, create Z and warn such person", an agent can speed up the first version of workflow.
Agencies and consultants who often deploy the same bricks
Versions, canvases and creation in natural language can help prototype several scenarios faster. But the agency must always document the dependencies, connected accounts, rights and consequences of an execution error.
Teams already equipped with MCP agents
The gain is greater if the team already works with assistants able to call tools via MCP. Zapier then becomes an execution layer that the agent can request, rather than a separate interface that must be opened for each change.
Less relevant if your priority is maximum hosting or control
If you are looking above all for self-hosting, detailed management of the infrastructure or a highly customized complex logic, n8n always deserves to be compared. If you prefer a very advanced visual editor to manipulate your scenarios, Make It also remains a reference to be examined.
Zapier, Make or n8n after this announcement?
| Priority | Option to watch first | Why? |
|---|---|---|
| Quickly create automation from an MCP-compatible agent | Zapier Next Gen Zaps | Workflow can be assembled and deployed from natural language interaction, with a return to Workflow Manager. |
| Visually build complex scenarios | Make | Its positioning remains very strong on visual representation and multi-stage scenario construction. |
| Technical control, extensibility and possibility of self-hosting | n8n | It is more appropriate for teams that want more control over execution and infrastructure. |
| Start with a simpler guided approach | Albato | Compare if you are looking for an accessible platform to connect applications without complex architecture. |
To go further, see our comparison Make vs n8n and our guide on automation of repetitive tasks in a small enterprise.
The limits to be checked before toggling
- Early access: functions, limitations and access modalities may evolve. Check the official documentation before starting a critical process on it.
- Cost of implementation: a simpler interface does not change the fact that tasks and certain AI operations are billed according to usage.
- Workflow Readability: reread the diagram and the connections created by the agent. Do not validate an automation that nobody on the team understands.
- Access rights: An agent that handles multiple applications can inherit sensitive connections. Limit permissions to what is strictly necessary.
- Error Management: Test incomplete cases, duplicates, empty fields, API quotas, and the unavailability of a third party service.
- Portability: Compare the cost of exit. Plus the workflow depends on Zapier-specific functions, plus migration to Make, n8n or an internal solution may require work.
Next Gen Zaps makes Zapier more appealing in an environment where AI agents become the main interface for work. However, the key selection criterion remains the same: Which tool does your process run in a reliable, understandable and economically sustainable way?
A signal to watch this week
Zapier organizes ZapConnect 2026 on 23 September 2026. The event is dedicated to automation by AI. As Next Gen Zaps has just been documented a few days before, new details may still appear. Logicelio will update this guide if Zapier confirms a change in availability, pricing or scope.
Frequently asked questions
What is Zapier Next Gen Zaps?
It is an early access automation system that allows to build and deploy Zapier workflows in natural language from an MCP-compatible client, while maintaining a visual manager and editing capabilities.
Is Next Gen Zaps available free of charge?
The Zapier documentation consulted the 21 September 2026 shows the Free plan as ineligible and the Professional, Team and Enterprise plans as eligible for early access on request.
Is it the same as Zapier MCP?
No. Zapier MCP refers to the ability to expose Zapier actions to MCP-compatible assistants. Next Gen Zaps uses this logic to go further and allow an agent to build and deploy complete workflows.
Should we abandon Make or n8n?
No. This announcement mainly improves Zapier's interface for creating and maintaining automations. The choice still depends on cost, integrations, governance, technical control and, for n8n, hosting.
Limit: Next Gen Zaps is in early access. Availability, quotas and pricing terms may change quickly. Check the Zapier documentation before making a purchase decision or migrating a critical process.
Product sources used
Analysis of 18 September 2026. Positioning and function information is verified primarily in the official resources of the vendors. Make: official source ; n8n: official source ; Albato: official source ; Editorial methodology Logicelio