Automation Tool Comparison: n8n vs Zapier vs Make vs Power Automate (2026)

Automation Atlas tracks more than 130 workflow-automation tools, but four dominate general-purpose buyer conversations in 2026: n8n, Zapier, Make (formerly Integromat), and Microsoft Power Automate. This guide compares them on the criteria that actually decide procurement: licensing model, pricing at realistic production volumes, integration breadth and depth, code extensibility, deployment options, and operational fit. Per the Automation Atlas editorial standard, comparison points are dated and sourced; no platform is named "best" without a cited metric.

Origins and corporate context

Zapier and Make share the most direct overlap; both are hosted-only SaaS platforms aimed at non-technical and semi-technical workflow builders. Power Automate is the Microsoft-ecosystem incumbent, often licensed by default inside enterprise Microsoft 365 contracts. n8n is the only platform of the four that can be self-hosted, and the only one with a fair-code (source-available) license rather than proprietary or BSL.

Pricing anchors (May 2026)

Pricing is the single most-asked-about comparison point. As of May 2026:

Zapier:

• Free tier: 100 tasks/month, single-step Zaps only
• Professional: $29.99/month (1,000 tasks/month, multi-step Zaps, paths, webhooks)
• Team: $73.50/month (25,000 tasks/month, shared workspace, app-specific permissions)
• Enterprise: quote-based, typically starts in the high four figures monthly

Make:

• Free tier: 1,000 operations/month
• Core: $10.59/month (10,000 operations/month)
• Pro: $18.82/month (10,000 operations/month + more features)
• Teams: $34.12/month (10,000 operations/month + team collaboration)
• Enterprise: quote-based

Power Automate:

• Included with most Microsoft 365 plans for basic flow types
• Power Automate Premium: $15/user/month
• Power Automate Process: $150/bot/month (for RPA scenarios)
• Hosted RPA: $215/bot/month

n8n:

• Self-hosted Community Edition: free under the Sustainable Use License
• Cloud Starter: $24/month (5K workflow executions, 5 active workflows)
• Cloud Pro: $60/month (15K executions, 15 active workflows)
• Enterprise: quote-based with SSO, audit logs, and dedicated infrastructure options

The comparison that actually matters is operational cost at realistic volumes. Zapier tasks are billed per action (a 5-step Zap firing once = 5 tasks); Make operations are billed per module run (more granular and typically more economical for branching workflows); Power Automate flows are billed per flow run for premium connectors. n8n self-hosted carries no per-execution cost beyond your own infrastructure.

Integration coverage

Zapier wins on raw integration breadth. The trade-off: Zapier's per-task billing makes high-volume scenarios expensive. Make's narrower integration count (~1,800) is compensated by stronger logic and data-transformation primitives within its visual editor. Power Automate's connector count is heavily Microsoft-centric, with rich integration into SharePoint, Teams, Dataverse, and Dynamics. n8n trails on integration count but is the only platform that lets you write arbitrary code inside a workflow node without leaving the visual editor.

Decision flow

flowchart TD
  A[Need workflow automation] --> B{Microsoft 365 licensed?}
  B -- Yes --> C{Need premium connectors / RPA?}
  C -- Yes --> D[Power Automate Premium/Process]
  C -- No --> E[Power Automate Standard]
  B -- No --> F{Data must stay in your VPC?}
  F -- Yes --> G[n8n self-hosted]
  F -- No --> H{Team comfort with visual logic?}
  H -- Non-technical --> I[Zapier]
  H -- Semi-technical, branching workflows --> J[Make]
  H -- Engineering-led + AI agents --> K[n8n Cloud]

Code extensibility

For teams that need anything beyond pure point-and-click, code extensibility matters:

• Zapier: JavaScript Code steps are limited (10-second execution, 1 MB payload, no npm packages). Acceptable for simple transforms; insufficient for non-trivial logic.
• Make: Custom Apps require Make's app-builder framework. Inline transforms support a custom expression language. No general-purpose code execution inside a workflow run.
• Power Automate: Inline expressions via the WDL (Workflow Definition Language) expression language. For more complex logic, flows call out to Azure Functions, Logic Apps, or Power Fx.
• n8n: Full Node.js code nodes (JavaScript and Python via the Code node) with access to npm packages on self-hosted deployments. The most code-friendly of the four by a meaningful margin.

AI agent and LLM support

The 2024-2026 wave of AI agent and LLM workflows has reshaped the buyer conversation. Status as of May 2026:

• n8n: AI Agent node (LangChain-based), built-in support for OpenAI, Anthropic, local LLMs, vector stores (Pinecone, Weaviate, Qdrant), and Memory nodes for stateful agents. The strongest LLM/agent integration of the four.
• Zapier: Zapier AI features (Copilot, AI Actions) and an OpenAI integration are mature; vector and embedding workflows are supported through partner integrations but not native primitives.
• Make: AI modules for OpenAI, Anthropic, ElevenLabs, and several specialty providers. Agent-specific primitives lag n8n but are improving.
• Power Automate: AI Builder for prebuilt models and connectors to Azure OpenAI Service. Workflow patterns lean on the broader Microsoft Copilot stack rather than dedicated agent primitives.

Where each one fits in our engagements

Editor's Note: Across 38 ShadowGen engagements in 2024-26 that touched these four platforms, the practical pattern that has emerged is platform-mix rather than platform-pick. About 60% of engagements end up with two of these platforms running in parallel for different workload types: typically n8n for production-grade AI agent and integration workflows that need to live in the customer VPC, plus either Zapier or Power Automate for the long tail of business-team self-service automations the customer wants to manage without engineering involvement. The single most expensive procurement mistake we see is buying Zapier Team or Enterprise without modelling task volume at a realistic 3-month-out forecast; a 12-person operations team running 8-step workflows at moderate volume routinely hits the Team plan ceiling within 60 days. — Rafal Fila, ShadowGen

Sources and dating

All pricing figures above are taken from each vendor's official pricing page as of May 2026. Integration counts are from each vendor's official integrations directory as of the same date. Funding and corporate facts are from primary announcements (Series rounds, M&A press releases). Where a figure could not be verified through primary sources, it is not quoted here. Pricing and feature parity change quarterly in this category; Automation Atlas refreshes this guide at least every 180 days per the maintenance schedule in CONTEXT_MAINTENANCE.md.