Why Your Zapier or Make Connection Keeps Disconnecting

The Reconnect Prompt That Won't Stop

Your lead workflow was working fine last Tuesday. Then Wednesday morning you open Zapier, or Make, and there it is again, a red warning icon next to your HubSpot connection, or your Gmail connection, or Twilio, telling you the account needs to reconnect before anything else runs. You click reconnect, log back in, everything goes green again, and you move on with your day. Then it happens again in three weeks. Or ten days. Or, if you're unlucky, it happens mid-week while a dozen leads are sitting in queue and nobody's watching the task history closely enough to notice.

This is one of the most common tickets we see when auditing an existing automation setup, and the root cause is almost always the same: a connection built without any plan for what happens when a token eventually expires. It's fixable, and once it's fixed properly it stays fixed.

Why the Reconnect Prompt Keeps Showing Up

Most of the apps you're connecting through Zapier or Make, Gmail, HubSpot, Google Sheets, Slack, don't use a simple username and password behind the scenes. They use OAuth, which hands the automation platform an access token and a refresh token instead of your actual login credentials. The access token is short-lived, usually an hour or less, and gets silently renewed in the background using the refresh token. You never see this part when it's working.

The problem is the refresh token isn't permanent either, and a handful of common, mostly unremarkable events can kill it without any warning to you or Zapier or Make. When that refresh fails, the platform has no choice but to ask you to log in again, because it genuinely no longer has anything valid to work with.

The Four Things That Actually Break a Connection

A handful of causes account for nearly every reconnect prompt we've traced during a client audit.

A password change on the connected account is the most obvious one. If someone resets their Gmail or HubSpot password, including a forced reset after a security alert from the platform itself, most services invalidate every existing token tied to that login. Zapier and Make aren't notified in advance. They just start failing the next time a refresh is attempted.

Google and Microsoft's own token expiration rules cause a lot of this too. Google's documentation lists a refresh token as no longer valid if it hasn't been used in around six months, and it also revokes tokens when certain account security settings change, like adding two-factor authentication. Microsoft 365 has similar behavior tied to conditional access policies. Neither company is targeting your workflow specifically, this is standard security practice across their platforms, but it means a Zap or scenario that only fires occasionally can lose its connection just from sitting idle for a stretch.

An admin revoking third-party app access is another one. In a Google Workspace or Microsoft 365 admin console, IT can review and revoke connected app permissions, sometimes as a routine security cleanup, sometimes because someone flagged an app they didn't recognize on a list. If the Zapier or Make connection wasn't clearly labeled when it was set up, it's an easy thing to revoke by mistake and a hard thing to trace back afterward.

The connection being tied to a person instead of the business is the one we see most, and it's also the most fixable. Someone set up the original Zap or scenario using their own personal Gmail or HubSpot login because it was the fastest way to get something working that afternoon. Then that person changes their password for reasons that have nothing to do with automation, goes on leave, changes roles, or leaves the company entirely, and the whole workflow goes down with them because nobody else even knew which login it depended on.

What Actually Breaks Downstream

The reconnect prompt itself is annoying but it rarely costs anything by itself. The real damage happens in the gap before someone notices the connection has dropped. In Zapier, a failed connection usually means every run for that Zap simply stops, no partial success, and the task history fills with error entries that most people aren't checking daily. In Make, a broken connection can pause the entire scenario depending on how the modules are wired together, one of several practical differences worth knowing before you pick between the two, which we've laid out alongside n8n. Either way, leads coming in during that window don't get logged in HubSpot, follow-up texts through Twilio don't go out, and support replies that were supposed to auto-send from a WhatsApp inbox just don't happen. It's a different failure mode from Zapier workflows outgrowing their own branching logic as volume climbs, but the two get mistaken for each other constantly since both show up as "leads randomly stopped arriving." Nobody gets an error notification that anyone actually sees, the automation just quietly stops doing its job. We've seen this run for over a week on a client's lead intake workflow before someone noticed conversion numbers looked off and started digging into why.

How We Set Up Connections So They Actually Hold

There's no version of OAuth where a token literally never expires. The realistic goal is to stop a reconnect prompt from silently breaking things, and to stop the whole setup from depending on one person's personal login. Here's roughly what we change when we take over an account that's had this problem repeatedly.

Move the connection to a dedicated business account instead of a person's inbox. For any app that requires an OAuth login (Gmail, HubSpot, Google Sheets), we set up a shared integration login, something like automations@yourcompany.com, that isn't tied to any one employee's password habits or their employment status. This one change fixes the most common root cause we run into, on its own.

Use API keys or private app tokens instead of OAuth wherever the platform allows it. HubSpot supports private app access tokens as an alternative to the standard OAuth flow, and those don't get killed by a password reset the same way. Twilio and the WhatsApp Business API both authenticate through API keys and long-lived tokens rather than a login session that can expire on you mid-workflow. Where the tool gives you the option, we take the token-based route over OAuth every time. It's just a sturdier foundation to build on.

Add a logging layer that doesn't depend on the connection most likely to break. We build a step early in the n8n or Make workflow that writes every incoming lead or event straight into a Google Sheet the moment it arrives, before it tries to reach HubSpot or fire off a Twilio message. If the HubSpot connection drops for whatever reason, every lead is still logged somewhere and nothing gets lost while someone reconnects.

Set up a failure notification someone will actually see. Zapier and Make both have built in failure alerts, but they usually land as an email in an inbox nobody checks in real time. We route connection failures to a Slack channel, or sometimes a WhatsApp message straight to whoever owns the workflow, so a broken connection gets caught in an hour instead of sitting there for a week.

Put a recurring check on the calendar. For tokens that expire on a roughly fixed schedule rather than a random event, we set a reminder to proactively re-authenticate ahead of the deadline instead of waiting for the failure to hit. It's a five minute task every few months and it avoids the surprise version of this entirely.

For a dental or medical clinic running appointment confirmations through Twilio, or a call center routing inbound leads by region, this kind of setup is the difference between a token expiring quietly on a Tuesday and a token expiring while nobody notices until Friday's numbers look wrong.

"Isn't This Just How OAuth Works? Nothing To Fix?"

Fair pushback, and it's partly true, honestly we would've said the same thing a few years back before we'd audited enough of these to see the pattern. OAuth tokens expire by design. That's normal, expected behavior built into how Google, Microsoft, and most connected apps handle security, and it isn't specific to Zapier or Make as platforms. We're not claiming you can build a connection that never asks to reconnect again. Anyone telling you that would be overselling it.

Most broken setups are missing anything built around the assumption that a token would eventually expire. Nobody planned for the reconnect, so when it happens the workflow just stops instead of degrading gracefully. Build the workflow so a token expiring is a small, contained event: an alert fires, someone reconnects in five minutes, and nothing gets lost in the meantime. Compare that to a workflow going dark for however long it takes someone to stumble across it, which is what's actually happening in most of the accounts we take over.

Signs Your Connection Setup Needs a Rebuild

A few patterns show up consistently across the accounts we've reviewed. The connection is logged in under a personal email address rather than a shared business one. Nobody on the team can say with confidence whose login the workflow is actually running through. You've had a reconnect prompt hit more than twice in the last quarter, and each time whoever happened to notice first just fixed it themselves, with no documented process behind any of it. Failure notifications land somewhere nobody checks daily, usually an inbox that already has a thousand unread emails in it. And there's no fallback logging layer, so a dropped connection means data is genuinely lost for that window rather than just delayed.

If two or more of those sound familiar, the setup is running on luck more than design, and it'll keep costing you leads or replies until someone rebuilds the connection layer properly.

Fix the Connection, Not Just the Symptom

Reconnecting the same Zap or scenario for the fifth time this year buys you a few more weeks at best, until the same token expires again for the same underlying reason nobody addressed. Once someone actually looks at why the connection keeps dropping instead of just clicking reconnect and moving on, the fix tends to be straightforward.

Book a free 60-minute call and we'll audit your current Zapier or Make connections, find out exactly why they keep dropping, and rebuild the setup so a token expiring is a minor inconvenience instead of a week of missed leads and support replies.