What did Relay do that its replacements may not?

Relay had four first-class human-in-the-loop step types — approvals, data input, task completion and manual path selection — with assignment to users, dynamic roles or Slack channels, interactive approval in Slack, due dates, reminders and configurable escalation. Most alternatives offer a single approval action. If your workflows relied on dynamic role assignment or per-step escalation policies, expect to rebuild that logic rather than port it.

Relay.app closes 23:59 PT, 14 September 2026

What happened to Relay.app?

The Relay team was acquihired by Google and joined the Chrome team; founder Jacob Bank became VP of Product for Chrome. The product was shut down rather than sold or transferred, which is why there is no continuity path, no acquirer running it in maintenance mode, and no "your account has moved to" email coming.

Date	What happened
16 July 2026	New signups and paid upgrades stopped
17 August 2026	Shutdown announced publicly
15 August 2026	Free accounts closed
14 September 2026, 23:59 PT	Paid accounts close. Workspace, account, credentials and tokens deleted.

Worth saying plainly, because it shapes how you should read the rest of this: Relay was good at the thing it was built for. It was not a weak product that lost. It had four first-class human-in-the-loop step types — approvals, data input, task completion and manual path selection — with assignment to individual users, dynamic roles or Slack channels, interactive approval inside Slack, due dates, reminders, and configurable escalation on a per-step basis. That is more granular than most of what remains. If your workflows leaned on that, you are not going to find a drop-in equivalent, and you should plan for rebuilding the approval logic rather than porting it.

What does the Relay.app export actually give you?

This is the part that determines how the next fortnight goes, and it is easy to discover too late.

Relay's own migration documentation is unusually candid: different workflow builders use different JSON structures, so the JSON export is not perfectly transferable to other tools without AI interpretation of the structure first, and it will not work through their native importers. There is no interchange format between automation platforms. This is not a Relay failing — it is true across the category, and it is worth internalising as a lesson about the next platform you pick.

Relay's actual recommended path was to export a prompt.md description of each workflow and have an AI rebuild it in the destination tool. That works better than it sounds, with one important caveat: a description round-trips the happy path and quietly loses the exception handling. Retry behaviour, what happens when an approval times out, which branch runs when a field is empty — these are the parts a natural-language summary flattens, and they are the parts that fail in production three weeks later.

Export both, and read the prompt.md

The JSON — records structure faithfully. Keep it as the reference of record even though nothing will import it.
The prompt.md — what the rebuild will actually be driven from. Read each one against the workflow it describes and add back what it omitted, especially error paths and approval escalation.

What is not exported at all

Execution history. The export covers definitions, not runs. If you need a record of what executed, when, with what result and who approved it — for an audit, a dispute, a tax enquiry, or any obligation with a retention period — you have to capture it manually, and you have until 14 September.

This is the single most commonly regretted omission in a platform shutdown, because the need for it appears months later, when the evidence is gone. If your workflows touched invoicing, payments, customer commitments or personal data, treat this as the first task rather than the last.

The twelve-day checklist

Inventory what is actually running. Every active workflow, its trigger, the systems it writes to, and who depends on the output. Expect to find automations nobody remembers building — those are the ones that break silently in October.
Export JSON and prompt.md for every workflow, including the ones you think you will retire. Deciding later is cheaper than recovering never.
Capture run history for anything with an evidential value. Screenshots are inelegant and they are better than nothing.
List every connected app. Credentials and tokens are deleted at shutdown, but the authorisations on the other side are not. Anything you do not migrate should be revoked in the source system, or you leave standing grants to a product that no longer exists.
Write down the approval logic separately from the workflow: who approves what, at which thresholds, with what escalation and timeout. This is the highest-value and most easily lost artefact.
Rebuild in consequence order — whatever moves money or touches customers first, not whatever is quickest to recreate.
Run old and new in parallel where the calendar allows it. It will not allow much.

Where should you move from Relay.app?

Relay's own documentation pointed to Zapier, Make (which offered white-glove migration) and n8n, plus AI assistants for one-off tasks. That is sound advice and you should weigh it on its merits rather than on ours.

If your constraint is…	Look at	Trade-off
Speed of rebuild, breadth of connectors	Zapier	Human-in-the-loop is now generally available, not beta, with approve/decline/change-data, email and Slack delivery, timeouts and reminders. Cloud-only — Zapier staff confirm there is no on-premises option.
Complex branching, ISO 27001	Make	SOC 2 Type II, SOC 3 and ISO 27001 at Enterprise, with Enterprise in an isolated AWS environment. Cloud-only.
Self-hosting, cost control	n8n	Genuinely self-hostable. First-class human-review steps across Slack, Teams, Telegram, WhatsApp, Gmail and Outlook. Note that its documentation does not describe approval audit logging — if you need that, verify it rather than assume it. Air-gapped licensing is an Enterprise arrangement; the Community edition runs offline without the licensed features.
Heavy enterprise orchestration	Camunda, and the BPM category	Camunda publishes an air-gapped installation guide. Much heavier to adopt; this is a platform decision, not a fortnight's migration.

A caution on how you evaluate: "governed automation" is now a crowded claim rather than an empty one. Zapier lists SAML SSO, SCIM, role-based access, audit logs, a queryable execution audit trail, log streaming to Datadog and Splunk, workspaces, action restrictions and app allowlisting. Workato has environment and project roles with an activity audit log and external streaming. Make has RBAC and ISO 27001. Anyone telling you that approval workflows or governance are novel is selling you something — including us, if we said it, which is why we are not.

Three questions worth asking while you are rebuilding anyway

A forced migration is an expensive way to get an opportunity, but it is one. You are about to touch every workflow you own. Three things are much cheaper to fix now than later.

Can you get your workflows out of the next platform?

There is no portability standard. The nearest thing is the Open Workflow Specification, a CNCF-governed effort formerly called Serverless Workflow — but its implementations are cloud-native developer orchestrators, not SaaS integration platforms, and no mainstream automation vendor has adopted it. BPMN 2.0 is the only broadly implemented portable notation, and only within the BPM world.

So the honest question is not "does this platform support a standard" — none of them do — but "what does its export contain, and have I tested restoring from it?" Ask before you commit, and test in the first month rather than the month you need it.

Would you be able to prove what ran?

You are currently discovering that Relay's execution history is not exportable. The same question applies to the destination: does it retain a record of executions and approvals, for how long, and can you get that record out? Approval checkpoints and audit trails are different things — a system can ask a human to click approve and retain nothing about what they were shown. See how to build an audit trail for what a defensible record actually contains.

Where does the processing happen?

This is the constraint that eliminates most options if it applies to you, and it applies more often than teams expect once regulated data enters a workflow. Zapier and Workato are cloud-only. Workato's On-Prem Agent is frequently misread: it creates a secure outbound connection from inside your network to Workato's cloud — the orchestration and processing still happen in Workato's cloud. Make ships on-prem agent installers with the same connectivity-bridge shape, though its documentation is thin on architecture. If you need processing inside your own boundary, the shortlist is short.

Frequently asked questions

When exactly does Relay.app shut down?

Paid accounts at 23:59 PT on 14 September 2026. Free accounts closed 15 August 2026. Signups stopped 16 July 2026. Workspace, account, credentials and tokens are deleted at the end of the shutdown period.

Can I import my workflows into Zapier or Make?

No. Relay's documentation states the JSON export will not work through other tools' native importers and needs AI interpretation first. Rebuild from the prompt.md export.

Why did it shut down?

Acquihire by Google; the team joined the Chrome organisation. The product was closed rather than transferred, so there is no successor.

What did Relay do better than its replacements?

Four distinct human-in-the-loop step types, dynamic role assignment, interactive Slack approvals and per-step escalation. Most alternatives offer one approval action. Plan to rebuild that logic.

Is my execution history exported?

No. Definitions only. Capture anything with evidential value before 14 September.

Should I move fast or move well?

Fast for anything currently load-bearing, then well. A rebuilt workflow that runs is worth more on 15 September than a well-architected one that is half-finished — but do not carry the shortcuts into December.

Where this leads

The uncomfortable lesson of a shutdown like this is not that the vendor failed. It is that the workflows were only ever legible inside one product, and when the product went, so did the definition, the history and the credentials. Portability, exportable evidence and processing locality are the three things that would have made this a weekend rather than a fortnight — and none of them are visible on a pricing page.

BarzelOps is built for that shape of problem: cross-system workflow automation with durable state on PostgreSQL, crash recovery and idempotency, human approval checkpoints with resumable execution, tenant isolation with signed evidence receipts, and a portable Governed Workflow Manifest intended to run in cloud, VPC, on-premises or air-gapped deployments. It handles Relay migration specifically, with dry-run validation and cutover reporting. To be clear about the term: Governed Workflow Manifest is our own name for a portable definition format, not an industry standard — no such standard exists, which is rather the point.

In practice

Automation you can leave, with the approvals and the evidence intact.

A workflow platform is only as safe as your ability to leave it. BarzelOps runs cross-system workflows with durable state, human checkpoints and signed evidence, exports them as a portable manifest, and migrates Relay workflows with a dry run before cutover.

In forceRelay.app closes at 23:59 PT on 14 September 2026

BarzelOps

Governed workflow automation across the systems that run the business.

Durable, idempotent execution: a timeout is retried once, never filed twice.
Human approval checkpoints that pause the workflow and resume it.
Isolation per entity or client, signed evidence receipts and a portable manifest; HubSpot, Xero, Gmail, Google Drive and Slack.

Free tier: 100 calls a dayPaid plans from $19 a monthLive on MCPize

Start free Ask by email Product page Documentation

BarzelVault

The AI action firewall: decide what an agent may do before it does it.

Approval thresholds and policy checks enforced before execution; human approvals that expire and escalate.
Cryptographically signed audit receipts: trigger, inputs, policy version, approver, outcome.
Credential isolation, spend and action limits, and an emergency kill switch.

Free tier: 10,000 calls a monthPaid plans from $199 a monthLive on MCPize