n8n vs Zapier vs Make at 100,000 Webhooks per Month
At 100,000 webhooks a month, self-hosted n8n on a $10 VPS costs roughly $10. Make at the same volume sits around $29 to $99 depending on operations per run. Zapier at the same volume lands between $73 and $799 depending on plan and task complexity. Volume is the lever that flips the answer; below 5,000 runs a month, the price difference rarely matters.
Quick clarifications
A webhook here means one inbound trigger that runs one workflow. A task in Zapier and an operation in Make are not the same thing; both bill per step inside a run, not per run. n8n self-hosted bills per server, not per step. These three units are why pricing comparisons online are usually misleading.
The bill that made me move everything
Around mid-2023 I had a client whose Zapier bill hit $487 in a single month. The workflow was simple: a webhook from a form fired six steps, including one OpenAI step. Their plan included multi-step Zaps but not the volume they had grown into. We migrated the same workflow to self-hosted n8n on a Hetzner CX22 over a weekend. Next month's bill: €4.51 for the VPS and $11 in OpenAI costs. Same workflow, same outcome. The migration paid for itself in 11 days. I have not put a high-volume workflow on Zapier since, though I still keep a free account for the things Zapier integrates with that nobody else does.
The opinion
Self-hosted n8n beats both Zapier and Make on cost past about 5,000 monthly runs. Below that, Zapier's polish and integration count make it the right pick for non-technical users. The mechanism: per-run pricing scales linearly with usage; a flat server cost does not. The cost of being wrong is either a surprise four-figure SaaS bill or a Saturday spent maintaining a server you did not need to spin up. Hold this loosely: if the only person who can fix the workflow is someone who has never touched Linux, self-hosting is a false economy.
Concrete prices at 100,000 monthly runs (as of October 2025)
- n8n self-hosted on Hetzner CX22 (2 vCPU, 4 GB RAM): €4.51 per month. Unlimited workflows, unlimited executions. Your time to maintain it is the hidden cost; for me, roughly 30 minutes a quarter.
- n8n Cloud Pro: $50 a month for up to 10,000 executions, then usage-based. At 100,000 runs you are looking at the Enterprise tier, which is custom-priced. (Source: n8n.io/pricing)
- Make Pro: $29 a month for 10,000 operations. At 100,000 operations you need the Teams plan at $99 a month. If each run uses three operations, 100k runs equals 300k operations, pushing you higher. (Source: make.com/en/pricing)
- Zapier Professional: $73.50 a month for 2,000 tasks; at 100,000 tasks you are on Team or Company plans. Pricing for 100k tasks/month on the public page in October 2025 is roughly $799 on the Team plan. (Source: zapier.com/pricing)
Where Zapier still earns its price
Per Zapier's app directory as of 2025, they list over 7,000 integrations. n8n has roughly 1,000 native ones plus the universal HTTP node. If your workflow needs to talk to some niche SaaS that nobody else has integrated, Zapier already wrote the connector. That convenience is real. If your workflow uses Slack, Google Sheets, OpenAI, Stripe, and Postgres, all three platforms support it natively and the price difference is the only meaningful axis.
Where Make has the edge
The visual flow in Make is, in my opinion, the best of the three for branching logic and aggregator patterns. The router and iterator modules feel more natural than n8n's IF and SplitInBatches. If your workflows are wide rather than deep, Make is genuinely pleasant to build in.
The hidden cost of self-hosting
I have spent roughly four hours total across the last twelve months keeping my self-hosted n8n running. That includes one upgrade that broke a custom node, one Postgres restart, and a TLS cert renewal that I had foolishly not automated. At the rate I charge clients, that is more than €4.51. Be honest about your own time before you celebrate the savings.
The 2026 starting prices, with units
As of mid-2026, per the pricing pages and the cipherprojects.com and automationatlas.io comparisons published that year: n8n Cloud Starter is $20 a month and meters in workflow executions (one execution covers the whole run regardless of node count), n8n self-hosted is free; Zapier Professional starts at $19.99 a month and meters in tasks (one task per action node that runs); Make Core starts at $9 a month and meters in operations (one operation per node call, so a workflow with ten nodes costs ten ops per run). The unit difference is the whole story. A workflow that does ten things costs you one execution on n8n, ten tasks on Zapier, and ten operations on Make. That ratio is why the 100k webhook scenario above shows the spread it does. Always price your specific workflow against the meter, not the headline number.
Frequently asked questions
Which platform is fastest?
Self-hosted n8n on a VPS in the same region as your other services is consistently the fastest because the network hops are shorter. Zapier and Make both run in shared infrastructure where queue times can spike under load.
Can I migrate workflows between them?
Not automatically. Each platform's node structure is too different for a clean export-import. You will rebuild by hand. Plan a half day per non-trivial workflow.
What about Pipedream and Activepieces?
Pipedream is excellent for code-first developers and competitive on price at low to medium volume. Activepieces is the open-source dark horse worth watching; the integration count is rising fast as of 2025.
Do I need a database with n8n?
Yes for production. n8n defaults to SQLite, which is fine for small workloads but will struggle past tens of thousands of executions a day. Point it at a Postgres instance and configure the queue mode. The n8n docs cover this in their scaling guide.
Is the n8n licence really free for commercial use?
The Sustainable Use Licence allows commercial use as long as the value is for you or your organisation, not for embedding n8n inside a product you sell. Read the licence yourself; it is short and clear, and the team updates it occasionally.
Pick by volume, not by marketing. At 500 runs a month, anything works. At 100,000, the server pays for itself before lunch.