There is no importer, so a Zapier or Make to n8n move is a rebuild. Export and inventory what you have, retire what does not run, rebuild the workflows that write to the CRM with lookup-before-create, run them in shadow mode next to the old ones, then cut over one trigger at a time and keep the old Zap or scenario paused for rollback.
Most RevOps teams outgrow Zapier or Make once their automations get complex enough: multi-branch enrichment, CRM write-backs with dedup logic, webhook chains between tools. At that point, n8n becomes a serious option because it meters by workflow execution, not by step. We wrote the full case for it in our n8n automation guide for B2B RevOps. This post covers the migration itself: how to get from Zapier or Make to n8n without breaking your CRM data or losing automations in the transition.
Should you migrate at all?
n8n wins when your workflows are multi-step, high-volume, and owned by someone comfortable reading JSON and the occasional Code node. The billing model is the clearest reason: n8n counts one execution per workflow run regardless of how many nodes it contains (n8n pricing). Zapier counts a task for every successful action step (Zapier pricing). Make counts credits per module run, with routers, filters and error handlers free (how features use credits). A workflow with one trigger and five action steps, running 2,000 times a month, is 10,000 tasks on Zapier but 2,000 executions on n8n.
Beyond billing, n8n gives you JavaScript and Python in Code nodes, branching logic, and the option to self-host in the EU for data residency.
Leave automations on Zapier or Make when they are simple one-trigger, one-action flows owned by non-technical people who maintain them without help. Zapier's library of pre-built templates and Make's visual scenario builder are genuinely easier for those cases. We compared the three tools in detail in our n8n vs Zapier vs Make comparison.
What carries over and what does not
Step 1: export and inventory
Zapier export. Go to Settings, then Security and data, then Data management. Click "Download my Zap workflows" to get your own Zaps as JSON, or "Download private and shared Zap workflows" if you are an owner or super admin and need everyone's. This feature is available on Team and Enterprise plans only (Zapier export docs). On Free or Professional, document each Zap from the editor (trigger, each step, field mappings) before you start.
Make export. Open any scenario in the Scenario Builder, click the three-dot menu, and choose Export blueprint. The download is a JSON file that includes modules, settings, and mapped values, but not connections or credentials. There is a 2 MB limit on import (Make blueprint docs).
Build an inventory spreadsheet. One row per Zap or scenario. Columns: name, trigger app, trigger event, action apps, run frequency (from the platform's analytics), last successful run date, and owner. Sort every row into one of three buckets:
- Rebuild in n8n: active, multi-step, writes to the CRM or handles revenue data.
- Merge: two or more Zaps that do related work and should become one n8n workflow with branches.
- Retire: has not run in the period your run history covers, duplicates another flow, or was a one-off experiment.
Group the rebuild list by trigger event. Workflows that share a trigger (for example, "new HubSpot deal") often share dedup and routing logic, so they migrate together.
Step 2: map Zapier and Make concepts to n8n nodes
n8n uses a different vocabulary. Here is how the core concepts translate. n8n processes data as arrays of items, where each item has a json key containing the payload. Every node receives and emits these items (n8n data structure docs).
Trigger (Zapier) or Trigger module (Make) maps to a Trigger node in n8n. Most app integrations have a dedicated trigger node. For custom events, use the Webhook node, which gives you a test URL in the editor and a production URL when the workflow is active. It supports GET, POST, PUT, PATCH, DELETE, and HEAD with optional Basic, Header, or JWT authentication.
Filter (Zapier) or Filter module (Make) maps to the IF node for binary conditions or the Switch node for multiple branches.
Paths (Zapier) or Router (Make) maps to the Switch node. n8n lets you define as many output branches as you need, each with its own condition.
Formatter (Zapier) or Tools module (Make) maps to the Set node for renaming and restructuring fields, or the Date & Time node for date formatting.
Code by Zapier or JavaScript module (Make) maps to the Code node, which runs JavaScript or Python and can process all items in a single run or one at a time.
Webhooks by Zapier or HTTP module (Make) maps to the HTTP Request node for outbound calls and the Webhook node for inbound.
Schedule (Zapier) or Schedule module (Make) maps to the Schedule Trigger node with cron expressions or simple intervals.
Delay (Zapier) or Sleep module (Make) maps to the Wait node.
Looping by Zapier or Iterator (Make) maps to n8n's native per-item processing (most nodes already run once per item) or the Split Out node for splitting an array; use Loop Over Items only when you need batches (for example to respect an API rate limit).
Sub-Zap (Zapier) or sub-scenario call (Make) maps to the Execute Sub-workflow node.
Error handling. Make attaches error handler routes to modules (Make error handling overview). n8n uses the Error Trigger node to build a separate error-handling workflow, plus per-node Retry On Fail and On Error settings (Stop Workflow, Continue, or Continue using error output) in the node settings.
Step 3: rebuild CRM-writing workflows so they cannot create duplicates
The highest-risk workflows in any migration are the ones that write to your CRM. A duplicated contact, a double-created deal, or a lost activity log can take hours to clean up and erode trust in the data. Build every CRM-writing workflow with these guardrails.
Lookup before create. Before any Create Contact or Create Deal node, add the HubSpot node's search operation or an HTTP Request node calling the CRM search endpoint to check whether the record already exists by email, domain, or a unique external ID. Only create if the lookup returns nothing. We covered this pattern in depth in our Clay to HubSpot write-back guide and the Clay, n8n, and HubSpot webhook handoff guide.
Idempotency key. Build a deterministic key from the source event's ID plus the action (for example "<eventId>:contact-upsert"), store the key of every processed event in an n8n data table or a database table, and skip any event whose key is already stored. That stops a retry or a re-sent webhook from writing twice.
Here is a small Code node that normalises the email, builds the key and decides whether to create, update or skip:
// Runs after the HubSpot search step. hubspotId is set when the contact already exists.
return $input.all().map((item) => {
const email = (item.json.email || '').toLowerCase().trim();
const eventId = item.json.eventId;
let action = 'skip';
if (email && eventId) {
action = item.json.hubspotId ? 'update' : 'create';
}
return {
json: {
...item.json,
email,
idempotencyKey: `${eventId}:contact-upsert`,
action,
},
};
});
Owner and routing in one place. Put your lead assignment logic in a single Switch or Code node so you can audit it. Do not scatter routing across three workflows. See our lead routing rules guide for the patterns that hold up at scale.
Consent and opt-out check. Before any outbound communication step, check the contact's opt-out or consent status. If your CRM stores GDPR preferences, query them. We documented the pattern in our GDPR opt-out sync guide.
Retries with backoff. Set each HTTP Request node that calls a CRM API to retry on 429 (rate limit) and 5xx errors with exponential backoff. For privately distributed apps (which is what a private app token in n8n uses), the burst limit is 100 requests per 10 seconds on Free and Starter and 190 on Professional and Enterprise, with daily limits of 250,000, 625,000 and 1,000,000 requests per account (HubSpot API usage guidelines). The page also notes the CRM Search API has its own stricter limits. Set your batch sizes to stay under your tier's ceiling.
Error Trigger workflow. Create a dedicated error-handling workflow that catches failures from any production workflow, logs the error details to a Google Sheet or Slack channel, and tags the failed item for manual review. This replaces Make's error handler routes and Zapier's Zap history checks with alerts you control.
Step 4: shadow run, then cut over
Never switch from the old tool to n8n in one move. Run them in parallel first.
Shadow mode. Build the n8n workflow to do everything the old one does, but instead of writing to the CRM, write to a staging table (a Google Sheet, an n8n data table, or a HubSpot sandbox). The shadow workflow logs what it would have written: record type, field values, and the action (create, update, skip). Compare these logs against the actual writes from the old Zap or scenario over a full business cycle. Differences reveal mapping errors, missing filters, or logic you forgot to port.
Cutover checklist:
- The shadow run has covered a full business cycle, including any monthly or quarterly jobs, and every difference is explained.
- List every system that sends to the old webhook URL.
- In a quiet window, pause the old Zap or scenario. Do not delete it.
- Switch the n8n workflow from staging writes to live CRM writes and activate it, so it uses its production webhook URL.
- Repoint every webhook sender to the n8n production URL.
- Watch the first executions in the n8n execution log and check the CRM for unexpected or duplicate records.
- Keep the old flow paused for a rollback window you agree up front; if something breaks, reactivate it and pause n8n.
- When the window closes, export any run history you need for an audit trail, then delete the old flow.
Cut over one trigger group at a time. If several workflows are triggered by HubSpot deal stage changes, migrate them together so you never have both the old tool and n8n writing to the same records from the same event.
The HubSpot workflow automation playbook and the AI agents for RevOps guide cover patterns for monitoring and testing these handoffs in production.
What it costs
Each tool meters differently, and the differences compound fast on multi-step workflows.
Zapier counts a task for every successful action step. Triggers and built-in tools (Formatter, Paths, Filter, Delay, Looping, Sub-Zap and others) do not count. A workflow with one trigger and five action steps, running 2,000 times a month, consumes 5 tasks per run = 10,000 tasks a month. Professional on annual billing at the 10,000-task tier is $129 a month (Zapier pricing).
Make counts credits per module run: trigger modules use 1 credit per run and action modules 1 credit per bundle processed, while routers, filters and error handlers are free (how features use credits). With one record per run, that same workflow is 6 credits per run (1 trigger + 5 actions) = 12,000 credits a month. Check the Make pricing page for the current price at that tier.
n8n Cloud counts one execution per workflow run, regardless of how many nodes the workflow contains. Those 2,000 runs are 2,000 executions. Starter is 20 euros a month billed annually for 2,500 workflow executions (n8n pricing). The self-hosted Community Edition is available on GitHub, so the cost there is your hosting and the time to run it (self-hosting guide).
This is example arithmetic at the time of writing, 20 Nov 2026. Check each vendor's pricing page for current figures.
Costs beyond the subscription. The rebuild time depends on how many workflows survive the inventory and how much logic sits in each. If you self-host n8n, add server costs (a small VPS or container) and the time to maintain updates. The team also needs to learn n8n's interface, which is more powerful but less polished than Zapier's for simple tasks. Factor all of this into your RevOps tech stack decisions.
Mistakes we see
In the migrations we run, the same handful of errors come up repeatedly.
Migrating everything at once. Teams try to rebuild the whole account in one sprint, skip shadow mode, and find CRM data issues weeks later. Migrate in small batches grouped by trigger, with a shadow period between each batch.
Forgetting to repoint webhook senders. The n8n workflow works, but the form or payment processor is still sending data to the old Zapier webhook URL. Keep a checklist of every external system that sends webhooks and update each one during cutover.
Ignoring the data model difference. n8n passes arrays of items between nodes. If a node upstream returns 50 items, the next node runs 50 times. Teams used to Zapier's one-item-per-execution model build workflows that accidentally process a batch as a single record, or process one record 50 times. Test with realistic data volumes, not a single test record.
Not testing rate limits. A workflow that calls the HubSpot API once per item in a 500-item batch will hit 429 errors within seconds. Use the Loop Over Items node with a batch size that keeps you under your HubSpot tier's limit, and configure retries on the HTTP Request node.
Both tools writing to the CRM during the parallel run. Shadow mode writes to a staging table until cutover. If the n8n workflow writes to the live CRM while the old Zap or scenario is still active, both tools create or update the same records and you end up with duplicates or conflicting field values.
One shared admin credential for everything. Use one credential per app with the narrowest scopes the workflow needs, owned by a service user, so one revoked login does not stop every workflow.
Skipping the error-handling workflow. In Zapier, failed runs show up in Zap history. In n8n, failures show in the execution log but nobody gets notified unless you build it. Set up an Error Trigger workflow on day one, before any production traffic.
Deleting the old flows too early. Keep the old Zap or scenario paused for the rollback window you agreed. The cost of keeping them paused is zero (no tasks consumed), and the rollback option is worth it.
Frequently asked questions
Can I import Zapier Zaps into n8n?
No. Zapier exports Zaps as JSON, and n8n can import workflows from JSON (n8n import/export docs), but the formats are completely different. There is no converter. Use the Zapier export as a spec document that describes what each step does, then rebuild the workflow manually in n8n. The exported JSON includes trigger types, action configurations, and field mappings, which saves you from having to open every Zap in the editor to remember what it does.
Can I import Make scenarios into n8n?
No. Make exports scenarios as blueprints in JSON format (Make blueprint docs), but n8n cannot read them. The same approach applies: export the blueprint, read it as a reference, and rebuild. Make blueprints include module configurations, connections between modules, and mapped values, but not credentials.
How long should Zapier and n8n run in parallel?
Run shadow mode (n8n logging intended writes without actually writing) for a full business cycle, long enough to include any monthly or quarterly jobs, until every difference is explained. After you switch to live writes on n8n, keep the old flow paused for an agreed rollback window before deleting it.
Is n8n cheaper than Zapier for RevOps workflows?
For multi-step workflows at moderate to high volume, almost always yes. n8n counts one execution per workflow run. Zapier counts a task per action step. A workflow with one trigger and five action steps at 2,000 runs costs 10,000 Zapier tasks but only 2,000 n8n executions. The gap widens as step count or volume grows. For simple two-step automations at low volume, Zapier's Free plan allows 100 tasks a month with two-step Zaps (Zapier pricing), which may be cheaper than any n8n plan.
Do my app connections carry over?
No. OAuth tokens and API keys are stored by the source platform and do not appear in any export. n8n's workflow export includes credential names and IDs but not the secret values (n8n import/export docs). You will need to create every app connection fresh in n8n: re-authorise OAuth flows, generate new API keys where needed, and test each connection before activating the workflow.
Should we self-host n8n or use n8n Cloud?
We usually suggest starting on n8n Cloud, which n8n hosts in Frankfurt for its hosted plans (n8n pricing), and moving to self-hosting when execution volume, network access or compliance requires it. Self-host if you need full control over where the data sits and who can reach the instance, or need n8n inside a private network. We covered the self-hosting decision in detail in our self-host n8n for EU and GDPR guide.
If you are planning a migration and want a second pair of eyes on the inventory, the rebuild plan, or the CRM safety checks, we run these as part of our automation practice and CRM and RevOps work. You can also ask us for a written second opinion on your current setup before committing to a move.