Relay.app Is Shutting Down: Why, Key Dates, Where to Move

The first deadline in the Relay.app shutdown has already passed. Free accounts and all their data were permanently deleted after August 15, 2026 at 23:59 PT. Paying customers have until September 14 at 23:59 PT, which is under a month from now, and then those accounts go the same way. New signups and free-to-paid upgrades have been off since July 16. If you run outbound or ops automation on Relay, the search for a Relay.app alternative is a migration project with a hard date on it, not a leisurely bake-off.

The announcement doesn’t say why, but the team was gracious about it with prorated refunds for annual customers, bonus steps and AI credits during the wind-down, and support through the transition. That’s a great way to close a product and we commend them for it. It still leaves you with the same problem though: workflows your business depends on need a new home in the next few weeks. This post is about picking that home well while you’re already doing the work of moving, and about what you can still salvage if the free deadline caught you.

Why Is Relay.app Shutting Down?

Relay hasn’t said. The announcement customers received on July 16, 2026 covers the wind-down logistics carefully, with the deadlines, the refunds, and the export instructions, but gives no reason for the decision. There’s no farewell post from the team as of this writing, so anything you read about the cause is guesswork, ours included.

What is public is the shape of the company. Relay was founded in July 2021 by Jacob Bank, who sold his previous startup Timeful to Google in 2015 and went on to lead product for Gmail and Google Calendar before leaving to start Relay. The company raised $8.1 million across two rounds, a $5 million seed led by Khosla Ventures in 2022 and a $3.1 million round led by Andreessen Horowitz in 2023, and pitched itself as a complete Zapier alternative with human-in-the-loop approvals built in. That’s a lean war chest for a head-on run at the automation incumbents, and the market has only grown more crowded since, but whether money, growth, or something else entirely drove the decision is not something anyone outside the company can confirm.

A month has gone by and there’s still no farewell post, so if you were waiting for an explanation before deciding anything, stop waiting. The honest answer to “why is Relay.app shutting down” is that they haven’t told us, and if they do we’ll update this post. The open question you have to resolve is where your workflows go next.

The Relay.app Shutdown Window, and What’s Left of It

The free half of the wind-down is over. Free users got 30 days from July 16, and after August 15 at 23:59 PT those accounts and their data were deleted. Paying customers are in the back half of a 60-day window: full access at no charge through September 14 at 23:59 PT, an automatic prorated refund on any unused annual time within five business days, and an extra 25,000 steps and 10,000 AI credits per month to ease the move. After September 14 the paid accounts and their data go too.

That leaves under a month, and the generous free access is what makes it tempting to file under “deal with later.” Later arrives fast when the thing you’re migrating is a set of live workflows with integrations, credentials, and run history behind them. A workflow you rebuild calmly in August is a workflow you’re not rebuilding in a panic in the second week of September. Treat what’s left of the grace period as time to migrate deliberately, not as time before you have to start.

What You Can Take With You, and What You Can’t

Relay made the exits super reasonable, which helps a lot. You can export your workflows, sequences, and MCP servers as JSON and as AI prompts, per Relay’s own export documentation. Your run history and your tables can be exported as CSV files. The AI-prompt export is useful, because it hands you a plain-language description of what each workflow does, which is exactly what you can feed a new builder to recreate it. Relay recommends that route over the raw JSON, and is blunt about why: different builders structure workflows differently, so the JSON “won’t work through their native importers” without an AI reading it first.

There’s a timing trap in the export itself that’s worth knowing about while you still have weeks rather than hours. A full workspace export doesn’t download on the spot. Relay generates the zip in the background, which their docs say can take up to 24 hours for a large workspace, then emails you a link that stays live for 48 hours. Start that job on September 14 and there’s a real chance it finishes into a dead account. Two other things to watch: workflow screenshots are generated separately and are not in the workspace zip, so if you want a visual record of how a flow was wired you have to grab those per workflow, and the run-history CSV caps at 100,000 rows per export and truncates past that.

You lose the plumbing though. When a Relay account is deleted, its stored credentials and tokens for your connected apps are deleted too, and Relay stops accessing those apps. So even a perfect JSON export doesn’t teleport a working workflow into another tool as you’ll need to reconnect every integration by hand on the other side like your Slack, your CRM, your email, and the rest. That reconnection work is unavoidable no matter where you land, which is the argument for choosing carefully. You’re forced to pay the switching cost once, so spend it moving to the platform you’d want for the next few years, not the one that most resembles the tool that’s closing.

There’s a file-by-file walkthrough in how to export your Relay.app data before it’s deleted, covering all four export types, a sequence that works if you’re squeezing it in between real work, and what each file is for once you’re rebuilding. Grab both workflow formats for every workflow, including the paused ones: once the account is gone you can’t come back for the one you skipped.

Once you have that AI-prompt export in hand, the rebuild doesn’t have to be yours to do. Paste one into the box below and we’ll rebuild it on Rills and send you a link to the finished thing.

If You Were on Relay’s Free Plan and Missed August 15

Then your account is gone and so is everything in it. Relay’s wording was that unexported customer content “will be permanently deleted at the end of the wind-down window,” and they’ve said nothing about restoring a closed account, so treat the files as unrecoverable. The usual escape hatch doesn’t exist either: paying to buy yourself into the longer window was never available, because free-to-paid upgrades were switched off on July 16, the same day the announcement went out. Anyone who found the email in a spam folder on August 16 had no move to make.

What you have left is more than nothing, though. A workflow is a description of a process, and the process still lives in places Relay didn’t control. The workflow’s own output is sitting in your Slack channels, your CRM records, your sent mail, and your spreadsheets, and reading a few weeks of that tells you what the thing did, on what trigger, with what content. Any screenshots you took while building are worth digging for. So is your own memory, which is better than you’d expect for a flow you looked at every week.

That’s enough to rebuild from. The free rebuild offer above takes a plain-language description of the workflow, not a file, so you can describe what it did in a paragraph and get back a working version without ever having had an export. It’s a worse starting point than a proper export, and we’d rather you had one, but it’s a long way from starting over.

What Relay Did Well, So You Don’t Lose It in the Move

Before you pick a replacement, be clear about why you were on Relay in the first place so you don’t drop the feature that mattered. Relay’s strengths were visual, multi-step workflows, bundled AI credits so model calls worked out of the box, and real human-in-the-loop steps. Those approval and manual-input steps are what set it apart from a fire-and-forget automation tool. You could put a person in the middle of a flow and make the automation wait for a decision.

If that approval step is the reason Relay earned a place in your stack, the three destinations you’ll hear about most each handle it differently, and none of them the way Relay did. Zapier’s Human in the Loop comes closest, since the reviewer can edit the drafted work before approving, but it’s a premium app starting on the Pro plan, every reviewer needs a Zapier account to respond, and each decision counts as a task. n8n gives you the machinery free on every plan, including self-hosted, but there’s no approvals inbox and the record of who approved what expires with the execution log. Make’s native approval app runs on Enterprise only, so everyone else splits the scenario in two, stashes the pending action somewhere, and emails themselves approve and reject links.

The through line is that you’d be leaving a tool that treated human review as a first-class step for tools that treat it as a plan tier or a hack you assemble yourself. The agent-first options land in a similar place for a different reason. Lindy will hand you a draft to approve, but the approving never gets any lighter, so the volume you review in month six is the volume you reviewed in week one. We ranked the main Relay.app alternatives by how well each one keeps that approval step, and compared five human-in-the-loop tools on where the approval actually lives, if you want the wider field before you commit.

Where Rills Fits as a Relay.app Alternative

Rills is built around that pause for review. It’s an approval layer for AI that acts on your company’s behalf: an AI proposes a consequential action, a human approves or rejects it from a mobile queue, and the workflow continues. If you came to Relay for human-in-the-loop steps, this is the same idea but taken further.

The review burden gets lighter as trust builds on Rills. Relay’s approval step is a switch that’s on or off per workflow and stays the same on the hundredth approval as on the first. Rills scores its confidence on every proposed action against the review step’s inputs. Patterns you reliably approve start clearing on their own, while anything unusual keeps stopping for a person. So a sequence sending renewal nudges to known contacts and cold first-touch emails to new domains can handle both on their own terms, waving the routine renewals through while the risky cold sends still wait for a look.

Every action leaves a decision record. When a prospect complains or a manager asks why a particular email went out, you can pull up what the AI proposed, who approved it, when, and the context they saw. That record lives on every plan with a generous 90 day minimum retention.

And the pricing doesn’t punish the thing you’re buying. On Rills, approvals and logic steps are free, and you’re billed only for high-value actions like an AI call or an external send. Adding a review step, a data transformation, or a branch that catches an edge case costs nothing. Relay metered steps and AI credits, so more workflow machinery meant a bigger bill. If the reason you’re adopting an approval layer is to review the risky actions, a model that charged you per review would work against you. There’s a full Rills vs Relay.app comparison if you want the side-by-side.

The last practical difference is where the approvals live. Relay’s human steps run inside a web flow, so approving means being at a browser. Rills puts the queue on your phone, so you swipe to approve from wherever you are, which matters a lot if you’re not always at your desk.

What Stops Rills From Shutting Down Too?

You’d be careless not to ask that a second time this year, so here’s the part you can check. Rills has taken no outside investment. No seed round, no board seat, no fund with an opinion about what this needs to be worth by 2029.

That changes which failure modes are on the table. Venture money buys speed, and the company owes an outcome large enough to return a fund in exchange. Under that math a product can be growing steadily, serving happy customers, and paying its own bills, and still be the wrong thing for the cap table to keep funding. We don’t know that this is what happened at Relay, and as noted above nobody outside the company does. The pressure just doesn’t apply to a company that never took the money. Rills runs lean and carries a small cost base, so the number it has to clear is customers covering the bill. Nobody outside can decide the product should be wound down because the curve isn’t steep enough.

Bootstrapped doesn’t mean permanent, though, and a vendor who tells you it does is selling a lie. A company without investors can run out of customers the same way a funded one runs out of runway. What it buys you is a slower, more predictable failure mode and no incentive to reprice you into an enterprise contract to make a quarterly number look right. Approvals and logic steps are free because that’s the product working as intended, not because a growth plan needs a loss leader for a while.

So judge us the way you should have been able to judge any vendor before this month: by the exits. Your workflows export as JSON and your table data as CSV whenever you want them, no support ticket and no wind-down announcement required. Data return and deletion terms are contractual in the Data Processing Agreement that applies to every account, not a policy page we can quietly revise. And if Rills ever did wind down, the floor is what Relay’s customers got: advance notice with full export access, never a surprise deletion. Migrating twice in one year is the thing you’re trying not to do again, so it’s fair to ask what the exit looks like before you carry your workflows in.

How to Move Your Workflows Before the Window Closes

The migration itself is straightforward if you do it in order. Start by exporting everything from Relay today, while your account is active: the JSON and AI-prompt versions of each workflow, your run history, and your tables as CSV. Kick off the full workspace export first, since that one runs in the background and can take a day to land in your inbox. Getting the data out has a hard deadline so do it now even if you want to do it again later.

Then rebuild the logic in Rills using the AI-prompt export as your source. Because that export already describes the workflow in plain language, you can hand it to Rills’s builder and get a working draft rather than starting from an empty canvas. Reconnect your integrations as you go, since those credentials never transfer between platforms for security purposes. As you place each step, put approval gates where the risk actually is like the external sends and consequential writes, and leave the internal logic to run untouched. Import your CSV tables where a workflow reads from them. Run each rebuilt workflow on a small batch before you point real volume at it, the same way you’d sanity-check any new automation.

Or skip that middle part entirely. Send us the AI-prompt exports and we do the rebuilding: a person here reads each one, builds the workflow on Rills, places the approval gates, and sends back a private link to the finished workflow with a short video explaining the choices we made. It costs nothing, needs no account, and the link works before you’ve signed up for anything, so you’re looking at your own automation running rather than at a marketing page. Send one workflow or send all of them. We’d rather spend the hours helping than watch another team lose theirs to a deletion.

None of this is a weekend if you have dozens of workflows, which is the case for spreading it across the weeks you have left instead of saving it for the end. There’s a step-by-step guide to switching from Relay.app if you want the checklist in one place.

Do the Move Now, Not on September 13

The mistake with a wind-down is treating the final date as the deadline. Your real deadline is whenever your team has the bandwidth to rebuild without cutting corners, and that window is the next couple of weeks, not the days before deletion. The free accounts that vanished on August 15 are the demonstration: a month of notice was plenty of time and still ran out for people who meant to get to it. Export your data today so it’s safe regardless, then rebuild at a sane pace while your Relay account still runs alongside as a reference.

If human approval is why you chose Relay, Rills is the closest thing to a pure upgrade of that idea, with the review shrinking as trust builds, a decision record on every action, and approvals that never cost you a credit. You’re going to rebuild these workflows once, so rebuild them somewhere the review gets lighter over time and the paper trail builds itself. Send us an export prompt and you don’t even have to do the rebuilding: we hand you back the finished workflow, and you decide from there. Or try a live demo first and swipe through a pending approval yourself.

Common questions

When is Relay.app shutting down?

Free accounts were deleted after August 15, 2026 at 23:59 PT. Paying customers keep access until September 14, 2026 at 23:59 PT, after which those accounts go too. New signups and free-to-paid upgrades have been off since July 16. Existing paid workflows keep running until the September date.

What happened to Relay.app?

Relay.app told customers on July 16, 2026 that it is winding down, without providing a reason. Free accounts closed on August 15, 2026 and paid accounts close on September 14, 2026, after which each account and its data are permanently deleted. New signups were turned off immediately, and annual customers get automatic prorated refunds.

I was on Relay's free plan and didn't export my data. Is it gone?

Yes. Free accounts and everything in them were permanently deleted after August 15, 2026 at 23:59 PT, and Relay has no archive to restore from. Paying to extend was never an option either, because free-to-paid upgrades were turned off on July 16. You can still rebuild a workflow from a written description of what it did, which is what the free rebuild offer runs on.

Can I export my workflows and data from Relay.app?

Yes, if you are on a paid account, until September 14, 2026. Relay lets you export workflows, sequences, and MCP servers as JSON and AI prompts, your run history, and your tables as CSV files. A full workspace export can take up to 24 hours to generate and the emailed download link lasts 48 hours, so start it well before your deadline.

What happens to my Relay.app data if I don't export it?

It is permanently deleted at the end of the wind-down window. Relay also deletes its stored credentials and tokens for your connected apps at that point, so it loses access to those apps too. Export anything you want to keep before the cutoff.

What is a good Relay.app alternative with human approval steps?

Rills is built around the approval step Relay users came for. Consequential AI actions wait in a mobile queue, a human approves or rejects, and every decision is logged. Approvals and logic steps are free, so adding review does not raise your bill.

Do I have to rebuild my Relay workflows from scratch to switch?

You rebuild the logic, but not from a blank page. Relay's JSON and AI-prompt exports describe each workflow well enough to hand to a builder, and you reconnect your integrations along the way since credentials do not transfer between platforms.

Can someone rebuild my Relay workflows for me?

Yes, and it costs nothing. Send Rills the AI-prompt export for a workflow and a person rebuilds it on Rills for you, using our agent, Ford. Once complete and verified, we'll send back a private link to the finished workflow plus a video walkthrough, usually within two business days. You can watch your own workflow run before creating an account.

Is Rills venture funded?

No. Rills has taken no outside investment, so there is no fund, board, or growth target deciding the product's future. The bar it has to clear is customers covering the bill, not a trajectory steep enough to justify a round. That makes it a slower company on purpose, and it removes the failure mode where a working product gets wound down for not scaling fast enough.

Relay had venture funding and still shut down. What if the next tool disappears too?

Judge a replacement by its exits, not its promises. Rills took no venture money, so no investor clock is running, but no vendor can promise permanence and you shouldn't accept one that tries. On Rills you can export your workflows as JSON and your table data as CSV at any time, and data return and deletion terms are contractual in the Data Processing Agreement on every account. If Rills ever wound down, you'd get what Relay's customers got at minimum: an advance-notice window with full export access, never a surprise deletion.