Relay.app Has Shut Down: How to Rebuild Your Workflows

Relay.app is no longer available. Its official shutdown notice confirms that the service closed in September 2026 and that accounts, workflows and run history have been permanently deleted. This is now a rebuilding guide, not an instruction to export from a working Relay dashboard.

Start with what you still control: a previously downloaded export, your own documentation, and records in connected apps. If you did not save an export, do not assume Relay can recover it. AgentLed cannot restore data deleted by another service.

Choose your starting point

Keep an untouched copy of any export. Work from a duplicate, identify sensitive fields, and do not paste API keys, tokens or complete customer datasets into a prompt or email.

Rebuild one useful workflow first

Choose a process that still has a clear owner and useful outcome. Record:

• What starts it and which inputs it needs.
• Which systems are read and which are changed.
• Which rules are deterministic and where an AI judgment is useful.
• What must be reviewed before a send, publication, payment or other consequential write.
• What a correct result looks like and when the process must stop.

An illustrative account-research workflow might read a company URL and approved product facts, produce a sourced summary, then prepare a follow-up draft. Creating that draft is separate from approving or sending it.

Use saved files as a specification, not a working import

If your export contains prompts, workflow JSON, tables or historical results, use them to explain the intended behavior. Check what is actually present; do not assume another tool accepts Relay's file format directly.

In AgentLed, describe the required output and the permitted actions, then review the proposed setup. This guide does not promise a native Relay importer, automatic conversion, or that a particular integration is already configured.

Reconnect each required app through the destination's supported connection flow. Start with read-only access where practical, and verify the exact account and workspace before enabling writes. Import saved tabular data only after checking fields, duplicates and access permissions.

Test without the old service

You can no longer rely on running Relay and its replacement side by side. Use saved outputs, source records and operator-approved examples as your reference.

1. Test with synthetic or minimized inputs and sends disabled.
2. Compare the output with a result the process owner has approved.
3. Check missing inputs, duplicate records, permission failures and retry behavior.
4. Verify that retries cannot create duplicate messages or records.
5. Approve one bounded live run before considering a recurring schedule.

A successful draft is not evidence that the send path, failure handling or schedule is ready.

When you do not have an export

Start from the business requirement instead of trying to reproduce an unknown configuration. Ask the process owner which result mattered, where its inputs came from, and what required approval. Mark gaps explicitly. Do not invent missing execution history or promise that deleted data can be recovered.

For help scoping a rebuild, contact AgentLed with a short description and redacted examples. Agree on scope, access, pricing and approval boundaries before sharing private exports or starting work. The former shutdown-period offer is no longer presented as available.

Rebuild checklist

• Preserve any files you already saved.
• Choose one process and name its owner.
• Specify inputs, outputs, permissions and stop conditions.
• Reconnect only the accounts it needs.
• Test drafts and failure cases before enabling writes.
• Review the first live result before scheduling further runs.

The goal is a verified replacement for one useful process, not an unsupported promise to recreate a deleted workspace.

Choose the operating model before the replacement tool

A migration is a chance to decide which steps should stay deterministic and which need context, memory, evaluation, or approval. If the old workflow should become more than a like-for-like rebuild, compare the alternatives to Zapier for agent-led operations before choosing the destination.