Relay.app shuts off paid accounts on September 14. The migration guides you have been reading mostly point you toward Zapier or Make, and for most workflows that is the right call. Both tools have good AI-assisted import flows and are strict supersets of Relay's API-based integrations.
But there is a category of Relay workflow that does not port. If any step in your automation touched a website that had no API (a supplier portal, a bank login screen, a government reporting tool, a legacy ERP), Zapier and Make hit a wall at exactly that step. They are server-to-server tools. Zapier's own staff say it plainly: "Zapier is back-end, server-to-server automation and browser work is out of scope."
Where Relay ran a browser
Relay's human-in-the-loop model let you insert steps where a person, or a Relay browser action, opened a portal and did something manually. Their own documentation metered "Browser automation" at 167 AI credits per minute. They carried five browser and scraping vendors as integrations (Apify, Axiom.ai, Browse.ai, Olostep, ScrapingBee). One of their published templates described scraping competitor sites using a stealth proxy to handle bot detection and waiting for full page loads.
The clearest signal is in the templates that ended in a human step. Relay's invoice processing template terminates with "Pay invoice ($)" because payment lives in a bank or vendor portal, and no payment portal publishes an API that a workflow tool can call. That step was always going to a person, or to a browser.
What each tool can rebuild
| Relay workflow type | Zapier | Make | WebRun |
|---|---|---|---|
| Trigger from an app with a public API | ✓ | ✓ | ✓ |
| Action that calls a public API | ✓ | ✓ | ✓ |
| Log into a web portal with no API | ✗ | ✗ | ✓ |
| Download a file from a login-gated page | ✗ | ✗ | ✓ |
| Fill a web form that blocks automated requests | ✗ | ✗ | ✓ |
| Scrape structured data with no export option | ✗ | ✗ | ✓ |
| Multi-step browser workflow: navigate, extract, submit | ✗ | ✗ | ✓ |
How WebRun handles the browser steps
WebRun runs AI agents in real cloud browser sessions. The agent navigates, logs in, clicks, reads, and returns data the same way a person would, without requiring the target site to publish an API. You describe the task in plain language. The agent handles the browser work and returns a result you can pass to the rest of your stack.
The agent opens the portal in a real browser session using your saved login, navigates to the report, and returns the file. No API needed on the supplier's end, no CSS selectors to maintain, no CAPTCHA wall.
For workflows that repeat on a schedule, WebRun's automation templates cover the common patterns: portal downloads, form submission, data extraction, and multi-step research. You can trigger any of them on a cron or from your existing orchestrator via API call.
The workflows most likely to break
The Relay workflows that will not survive a move to Zapier or Make are the ones that:
- Accessed an internal or legacy web tool with no public API (HRIS portals, procurement systems, supplier dashboards, government filing tools)
- Had a human-in-the-loop step where someone opened a browser to do something manually
- Pulled data from a site with bot detection, CAPTCHA, or IP restrictions
- Combined API steps with browser steps in the same sequence
WebRun handles the steps Zapier and Make cannot reach. Everything else in your stack stays the same.
Relay.app paid access ends September 14, 2026. To get started with browser automation, sign up at webrun.ai.
