Relay.app is shutting down. Here is where your workflows should go.
Relay announced on 16 July 2026 that it is winding down. New signups are already closed, free accounts end on 15 August and paid accounts end on 14 September. So the useful question is no longer how n8n compares to Relay feature by feature. It is what you do in the next few weeks. This page covers what to export before you lose access, what each Relay feature maps to in n8n, and where n8n is honestly the wrong answer for you.
Last updated: July 2026 · by Abhiman Sundararajan, Verified n8n Creator (Level 2)
Everything keeps running until your date. Exports stop being possible once the account closes.
What actually happened
Relay sent a service announcement to customers on 16 July 2026 saying the product is winding down. New signups and free-to-paid upgrades were switched off immediately. Existing workflows keep running until the relevant cutoff, so nothing breaks the moment you read this. Annual plans are refundable.
Worth saying plainly: Relay was a good product. The approval steps were better thought through than most competitors, the canvas was clean, and plenty of teams built real operations on it. Companies close for all sorts of reasons that have nothing to do with the quality of the software. If you are annoyed about having to rebuild, that is a fair reaction.
The practical takeaway is that you have weeks, not months, and the export window closes with your account.
Export everything first, before you pick anything
Do this today, before you evaluate a single replacement. Everything else on this page can wait a week. This cannot.
- Workflows, Sequences, and MCP servers export as JSON
- Tables export as CSV
- Run history and AI prompts are exportable too, and the run history is more useful than people expect for reconstructing edge cases
- A full workspace export can take up to 24 hours to generate, and the download link is only valid for 48 hours once it lands
That 24 hour build time and 48 hour link expiry is the trap. If you request the export on a Friday afternoon and do not check your email until Monday, you request it again. Start it now, and actually download the file to somewhere permanent rather than leaving it in an inbox.
One expectation to set: the JSON does not import into n8n, Zapier, or Make. None of them read Relay's format and no reliable converter exists. What the export gives you is an exact record of every trigger, branch, and field mapping you built, which turns the rebuild from remembering into transcribing. That is worth a great deal, just not the thing people hope it is.
Every Relay feature, and what it becomes in n8n
| In Relay | In n8n |
|---|---|
| Steps billing | Executions. One workflow run counts once no matter how many nodes it touches, and self-hosted has no meter at all |
| AI credits | Your own Anthropic, OpenAI, or Google key. You pay the provider directly at cost |
| Human-in-the-loop approvals | Send and Wait for Response on Slack, Send and Wait for Approval on Gmail, or the Wait node with a webhook |
| Paths | If, Switch, and Filter nodes |
| Sequences | Sub-workflows called through the Execute Workflow node |
| Tables | Notion, Airtable, Postgres, or Supabase. The data lives in a real database you own |
| MCP servers | Native MCP client and MCP server trigger nodes |
| Agents | The AI Agent node, with tools, memory, and your choice of model |
| ~200 integrations | ~500 native nodes plus any REST or GraphQL API through the HTTP Request node |
| Cloud only | n8n Cloud, or self-hosted on your own server, VPS, or Docker |
The billing line is the one worth pausing on. Relay charged per step, so a ten step workflow cost ten steps every run. n8n charges per execution, so that same workflow costs one. Self-hosted it costs nothing beyond the server. For anyone who was watching their Relay step count, the replacement is likely cheaper rather than a like-for-like swap.
What n8n gives you that Relay never did
- Self-hosting. Your workflows run on your server, your data never leaves it, and no vendor decision can switch them off
- JavaScript and Python code nodes, so logic that no prebuilt block covers is a few lines instead of a workaround
- Real error handling. Dedicated error workflows, per-node retry counts and delays, and failure branching so one bad record does not stop the batch
- Model choice. Swap between Claude, GPT, Gemini, or a local model without leaving the workflow
- A fair-code licence. The source is public, you can read it, and you can run it yourself indefinitely
That last one is doing more work than usual this month. The reason to care about self-hosting is not ideology, it is that a workflow running on your own server cannot be discontinued by an email.
What you give up moving to n8n
Every migration page tells you the destination is better in every way. This one will not, because you are going to find these out in week one anyway.
- Approvals take assembly. Relay's human-in-the-loop steps were genuinely well designed and worked out of the box. n8n gets you there with Send and Wait for Response on Slack, Send and Wait for Approval on Gmail, or a Wait node and a webhook. Same outcome, more wiring.
- AI credits become an API key. Relay handed you a credit pool and one bill. n8n asks you to bring an Anthropic or OpenAI key. Cheaper at volume and more flexible, but it is another account, another key to rotate, and a bill that moves with usage.
- There is no Tables. Relay had a database inside the product. n8n does not, by design. You point it at Notion, Airtable, Postgres, or Supabase. Better long term because the data is somewhere real, more setup on day one.
- The first hour is steeper. Relay optimised for feeling easy. n8n optimises for not hitting a wall. That trade is worth it if you are building something you will maintain for years, and it is not worth it if you have three simple workflows and no appetite for learning a tool.
Which replacement actually fits you?
n8n is not the right answer for everyone leaving Relay, and pretending otherwise would waste your time. Here are three honest profiles.
You do not want to be here again in two years
Relay closing is the second time a tool you depend on changed the terms underneath you. You want automations you can host yourself, workflows you can read and version like code, and a licence that does not depend on one company staying solvent. Someone on the team is comfortable with a server.
You need working automations back, fast
Nobody on the team is technical, the workflows were straightforward, and the priority is having things running again before the deadline rather than owning the stack. Your apps are mainstream ones with prebuilt connectors everywhere.
You want a visual canvas and a small bill
You liked building on a canvas, your workflows have real branching in them, and cost matters more than self-hosting or code. You want something managed, visual, and cheaper per run than the enterprise options.
If you are torn between the first two, read n8n vs Zapier. If it is between n8n and Make, read n8n vs Make. Both go through the cost models properly.
Where I can help
Rebuilding a set of workflows under a deadline is exactly the kind of work we do, and the deadline here is real. I am a Verified n8n Creator at Level 2. If you send me your Relay export, a free 30-minute call gets you a straight assessment: what rebuilds cleanly, what needs rethinking, what it costs, and whether n8n is even the right destination for your situation.
Common questions.
When exactly does Relay.app shut down?
Relay announced the wind-down on 16 July 2026. New signups and free-to-paid upgrades are already switched off. Free accounts stop working on 15 August 2026 and paid accounts stop on 14 September 2026. Until your date arrives everything keeps running normally, so your workflows do not break early. The date that matters most is not the shutdown though, it is the day you export, because you cannot export anything after the account closes.
Can I import my Relay workflows into n8n?
No, and be careful of anyone who tells you otherwise. Relay exports workflows, Sequences, and MCP servers as JSON, but that JSON describes Relay's internal model, not n8n's. No tool converts one into the other cleanly. What the export is genuinely good for is documentation: it is a precise record of every trigger, every branch, and every field mapping you built, which turns a rebuild from guesswork into transcription. Export first, rebuild second.
What replaces Relay's AI credits in n8n?
Your own API key. Relay bundled AI usage into a credit pool you topped up through them. In n8n you connect an Anthropic, OpenAI, or Google key directly and pay that provider at their rates with no markup in between. It is cheaper once you run any real volume and you can switch models whenever you want. The cost is one more account to set up and a bill that varies with usage instead of sitting in a neat monthly bucket.
Does n8n do human-in-the-loop approvals like Relay?
Yes, though you assemble it rather than getting it prebuilt. n8n has Send and Wait for Response on Slack, Send and Wait for Approval on Gmail, and a general Wait node that pauses a workflow until a webhook fires. That covers approve or decline buttons in Slack, approval by email, or a custom form. Relay's version was more polished out of the box, and this is one of the few places where the move genuinely costs you some convenience.
Is n8n harder to use than Relay was?
For the first workflow, yes. Relay was built to feel approachable and it succeeded at that. n8n shows you more of the machinery up front, so expect the first hour to be slower and expect to learn how expressions and data pinning work. After that the curve flattens and the ceiling is much higher. If nobody on your team is technical and nobody wants to become technical, be honest about that when you pick, because it points at Zapier instead.
Should I move to n8n, Zapier, or Make?
Move to n8n if someone on the team is technical, you want to self-host, or you never want a vendor to be able to do this to you again. Move to Zapier if you want the shortest path back to working automations and your apps are mainstream. Move to Make if you want a polished visual canvas and your budget is tight. All three are reasonable. The wrong answer is waiting until September and then rebuilding under pressure.
Can I get a refund on my Relay annual plan?
Yes. Relay confirmed refunds are available for annual plans as part of the wind-down. Sort that out at the same time as your export rather than leaving it, because support capacity at a company that is closing tends to get thinner as the date approaches, not better.
Losing Relay? Let us map the move.
Pick a time. Bring your export or just a list of what you built. In 30 minutes you get a clear plan for rebuilding it, an honest read on which tool you should land on, and a price if you want it done for you. Free, no obligation.