Moving off Relay.app: Where Your Workflows Should Go, and How to Get Them There

A practical migration guide for Relay.app users — by Effibotics

Relay.app is winding down. Free accounts and their data are deleted on Saturday, August 15, 2026. Paid customers keep access through Monday, September 14, 2026. The data export window is open now.

If you built real operations on Relay — approval flows, AI steps, multi-app coordination — this is disruptive. But the easy mistake isn't picking the wrong tool. It's rebuilding the same brittle setup somewhere else without asking where the work should actually live. This guide covers both: where to go, and how to move.

Step 0: Export your data now (before you decide anything)

Do this today, regardless of where you land. Free-tier data is deleted first and there is no recovering it after August 15. Even if you're a paid customer with time until September, export now so a missed date can't cost you your workflow history.

While you export, inventory each workflow. For every one, write down:

• The trigger (what starts it)
• Every step, in order, and which app it touches
• Any AI steps — and copy the prompt text out verbatim
• Every human approval checkpoint
• How data maps from one step to the next

This inventory is the actual migration plan. You'll rebuild from it.

The four places Relay users realistically go

You chose Relay for four things most tools don't do well together: human approval steps, AI-agent-first design, built-in collaboration, and predictable pricing. That's the lens for judging every destination — not features in the abstract, but which of those you keep.

n8n — the closest match to everything you valued, minus the ease. Native approval steps, real AI agents, predictable cost. Self-host it and no vendor can wind it down. For someone who just lived through a shutdown, that last point is the whole argument. The one real cost is that it's the most technical to stand up and maintain.

Make — the visual builder that feels most like Relay, and cheaper than Zapier. Good if your workflows were more about connecting apps than running AI. But there's no native human-in-the-loop, AI is bolted on, and per-operation pricing gets fiddly on AI-heavy flows — the exact thing Relay's pricing spared you.

Zapier — the fastest, lowest-friction landing spot, with the biggest integration library anywhere. If your Relay workflows were simple and linear, this gets you live again this week. The catch: the two things Relay users loved most, approvals and native AI, are Zapier's weakest areas, and task-based pricing punishes AI-heavy or high-volume work.

Gumloop — spiritually the closest successor: agent-first, human review steps, easy to learn. If it was the AI and approval flow you loved, this feels the most like home. Be honest about one thing, though — it's a young, venture-backed startup, the same profile as the platform that just folded under you. Good product, real continuity risk.

Two more worth knowing by name: Activepieces (open-source, self-hostable, has approval steps — a softer landing for the ownership crowd) and Lindy (pure AI-agent play, if your Relay use was mostly the mini-agents).

The one question that decides it

Everything above collapses into a single fork: can you (or someone you'll pay) run something technical?

• If yes, and you never want to be shut down again → n8n, self-hosted.
• If no, and you want simple and fast → Zapier.
• If no, but the AI and approvals mattered most → Gumloop, eyes open on the startup risk.

How migration actually works

There is no automated importer. Relay does not export into n8n, Zapier, or anyone else's format. Every migration is a manual rebuild from your inventory. That's not a warning so much as a scoping reality — plan the hours, or have it built for you.

The process is the same shape wherever you land:

1. Inventory every workflow from your Step 0 export.
2. Rebuild each one in the new tool using the concept maps below.
3. Reconnect credentials — every app needs fresh authentication in the new platform.
4. Run both in parallel for a few days, compare the outputs, then cut over. Never hard-switch a live workflow.

Relay → n8n

The two rows that matter most coming from Relay are human-in-the-loop and sequences — n8n reproduces both cleanly. Your approval steps become the Send-and-Wait node, with approve/reject buttons or an editable form, and a wait-time limit, on whichever channel your team already uses.

Relay → Zapier

The rows to watch on Zapier are human-in-the-loop and sequences: both are fudged rather than reproduced. If approvals were core to how you worked, budget extra time here, or reconsider the destination.

What surprises people mid-migration

• Your AI credits vanish. In n8n you bring your own OpenAI or Anthropic key — often cheaper than Relay's managed credits. In Zapier, AI is metered per task. The prompts themselves port over by copy-paste with no rework.
• Pricing won't map one-to-one. Relay's steps-plus-credits model doesn't translate cleanly. Set expectations: n8n is flat or free self-hosted, Zapier charges per task, Make charges per operation. On high-volume or AI-heavy work, the gap between these is large.
• "Linked objects" have no direct equal. Relay's automatic pull of related CRM records (the deal, plus its company and contacts) becomes a manual lookup step in every other tool. Budget for it.
• Niche integrations may need rebuilding. Anything you ran through Relay via webhooks will need re-wiring, and coverage differs tool to tool.
• Collaboration and roles differ by tier. If your Relay setup was multi-user, check how the destination handles shared access before you commit.

Before you rebuild the same thing somewhere else

A shutdown forces a question most teams never get to ask on purpose: is this workflow even shaped right?

Plenty of Relay workflows grew by accretion — a step here, an approval there, a fix for a thing that broke once. Rebuilding them verbatim just moves the drag to a new address. The better move is to map where coordination actually fails first, then deploy only what earns its place, on infrastructure you own.

→ effibotics.com/booking