Prerequisites

• Zapier admin access (you need to view all Zaps and their connected accounts).
• A self-hosted n8n instance — see /zapier/ for the 10-minute docker-compose recipe and ~$5-20/mo VPS sizing depending on workflow volume.
• n8n's queue mode (Redis-backed) if you'll run more than a few hundred workflow executions per day — set `EXECUTIONS_MODE=queue` and add a Redis container.
• API credentials for each integrated service (Slack bot token, Notion integration token, Airtable PAT, Google service account, etc.) — Zapier abstracts these via OAuth, n8n re-OAuths or accepts API keys directly.
• A list of every active Zap, ranked by mission-criticality — the top 20% you'll port first.

Step 1 — Export from Zapier

1. Inventory your Zaps

   Zapier dashboard → Zaps. Filter by 'On' to see active workflows. Note for each: trigger app + event, action app(s) + action(s), filter steps, formatter steps, paths (branches), tasks consumed/month. Tasks/month is your cost driver — high-volume Zaps are the migration priority.

2. Document each Zap as a flowchart

   For each Zap, screenshot the editor view or write the steps as: `[Trigger: New row in Airtable / Table=Leads] → [Filter: Status = 'New'] → [Action: Slack post in #sales] → [Action: Send Gmail to lead.email]`. This is the spec for the n8n rebuild — n8n's node graph maps near 1:1.

3. Capture filter and formatter logic

   Zapier's Filter and Formatter steps have specific operators (Date/Time → Format, Numbers → Spreadsheet-style Formula, Text → Replace). Save the exact configurations — these become n8n's IF, Set, and Code nodes.

4. List connected accounts

   Zapier → Apps. List each connected account (Slack workspace name, Google account email, Airtable workspace). For each, you'll re-authenticate or generate fresh API credentials in n8n. Some apps (Notion, Airtable) need a fresh integration token from the source app's developer settings; OAuth-based apps (Slack, Google) re-OAuth in the n8n UI.

Step 2 — Import into n8n

5. Stand up n8n with persistent Postgres

   Don't use the default SQLite for production. docker-compose with: n8n + Postgres + (Redis for queue mode). Set `DB_TYPE=postgresdb`, `DB_POSTGRESDB_HOST=...`. Sign in at the n8n URL; first user becomes owner.

6. Add credentials for every integration

   n8n → Credentials → New. For each service in your Zap inventory: paste the API key or run the OAuth flow. n8n stores credentials encrypted with the `N8N_ENCRYPTION_KEY` env var — set this to a random 32-char value at install time and back it up; losing it means re-authing everything.

7. Recreate one Zap, end-to-end

   Pick the simplest Zap. n8n → Workflows → New. Add a trigger node matching the Zapier trigger (e.g. `Airtable Trigger` for new row). Add the action node (e.g. `Slack` → Post Message). Wire them. Test with a sample row. Compare output to what Zapier produces — confirm the field mapping matches before moving on.

8. Translate filter/formatter steps

   Zapier Filter → n8n IF node. Zapier Formatter → n8n Set node (for renaming/restructuring) or Code node (for arbitrary JS — e.g. `return [{json: {date: new Date(items[0].json.timestamp).toISOString()}}]`). Test each branch.

9. Handle webhooks and HTTP

   Zapier Webhooks (Catch Hook) → n8n Webhook node. n8n gives you a unique URL per workflow. Update the source service to POST to the n8n URL instead. The payload is the same shape — n8n exposes it as the trigger's output.

10. Activate and watch

    n8n workflow → Active toggle. Send a real production event. Watch n8n → Executions for the run. Compare the output to what Zapier did. Once stable, deactivate the corresponding Zapier Zap.

11. Migrate the rest in waves

    Top-20%-by-volume first. Then the long tail. Some Zaps will require a node that's not in n8n's catalog — write a custom node (TypeScript, ~50 lines) or call the third-party API directly via the HTTP Request node.

Field / concept mapping

Downtime estimate

Zero downtime via dual-run: keep Zapier active while you rebuild and test. Cutover per workflow is instant — flip Zapier off, n8n is already on. Per-workflow rebuild: 5-30 minutes for a simple 2-3 step Zap; 1-2 hours for a complex Zap with paths, filters, and custom formatters. A full migration of 50 Zaps is typically 1-2 weeks of part-time work spread over the migrator's calendar.

Common gotchas

• n8n's license is the Sustainable Use License — fair-code, not OSI. Free for internal business use including self-host; restricts re-selling n8n as a SaaS. If you need pure OSI/MIT, evaluate Activepieces (also covered on /zapier/) instead.
• n8n credentials are encrypted with `N8N_ENCRYPTION_KEY`. Losing this env var bricks all stored credentials — back it up the moment you set it.
• Default n8n SQLite is fine for testing, painful for production (executions table grows fast). Switch to Postgres at install time.
• Some Zapier connectors are wrappers around scraped APIs (e.g. some legacy CRM connectors). n8n doesn't have these — you'll fall back to the HTTP Request node + the official API.
• Zapier's task counter and n8n's execution counter measure differently. n8n is unmetered (your VPS is the cost ceiling), but high-volume polling triggers can max out a small VPS. Move to webhook-based triggers where possible.

Rollback plan

Per-workflow rollback is a single click — re-enable the Zapier Zap, disable the n8n workflow. Zapier holds onto Zaps for 90 days after deactivation, so if you cancel the Zapier subscription too quickly there's a re-activation grace period. Keep Zapier paid for at least the duration of the migration window, then a 30-day buffer post-cutover to catch any workflows whose translation has subtle bugs.