NewPlain-language diffs for n8n, Zapier & Make

The workflow was fine yesterday. Today it isn't. Here's why.

Something changed: a node got edited, a field got renamed, an API started returning a different shape, and now a workflow that ran for months without a problem is failing silently or doing the wrong thing. Nobody remembers making the change. The error message doesn't say which version broke it.

  • No credit card
  • Bring your own AI key
  • Fixes never auto-apply
Works alongside
n8nZapierMake
Cloud or self-hosted
Payment webhook → Order fulfillmentn8n
v41 · last good→v42 · today
Webhook
HTTP Request 3CHANGED
IF paid?
Fulfill order
Notify Slack
Versions
v38v39v40v41v42
Error handling removedHTTP Request 3 · 3 parameters

This change removed the retry on your payment webhook, which means a failed charge no longer gets retried before the customer sees an error.

Proposed fix

+ retryOnFail: true

+ maxTries: 3

+ waitBetweenTries: 1000

Nothing applies until you approve.
retryRetry removed on payment webhook
renameField email renamed to contact_email
credentialSlack node no longer references its credential
rewireIF branch rewired to skip Fulfill order
retryRetry removed on payment webhook
renameField email renamed to contact_email
credentialSlack node no longer references its credential
rewireIF branch rewired to skip Fulfill order

FlowLens keeps a version history of your n8n, Zapier, and Make workflows, diffs them automatically, and explains in plain language what changed and why it matters.

Not

"Node 'HTTP Request 3' modified."

Something closer to

this change removed the retry on your payment webhook, which means a failed charge no longer gets retried before the customer sees an error.

The actual problem

The actual problem (not the abstract one)

Most teams don't find out a workflow broke until someone downstream complains. A Slack message doesn't go out. An order doesn't get fulfilled. A lead doesn't land in the CRM. By the time anyone opens the workflow to look, they're staring at a JSON tree with forty nodes, no idea which one changed, and no record of what it looked like last week.

n8n, Zapier, and Make all show you that a run failed. None of them tell you what changed since the last time it worked, and reconstructing that by hand, node by node, is the part that actually eats an afternoon.

Downstream

A Slack message doesn't go out.

Downstream

An order doesn't get fulfilled.

Downstream

A lead doesn't land in the CRM.

Then someone opens the workflow

What your platform shows

✕ Execution failed

What you actually need

What changed since it last worked

How it works

How FlowLens actually works

Four steps, from the moment a workflow changes to a fix you've reviewed.

It takes a snapshot every time your workflow changes or runs.

01

Import a workflow once, and FlowLens keeps a timestamped history, not just the current state, every prior version.

HistoryPayment webhook
v42on changetoday 14:01
v41on runOct 18 09:12
v40on changeOct 12 16:40

It diffs two versions and tells you what moved.

02

Added nodes, removed nodes, changed parameters, rewired connections, structurally, not just "something's different."

+Added nodes
−Removed nodes
~Changed parameters
↻Rewired connections
The part that matters✦

It explains the diff in plain language.

Instead of a raw list of field changes, FlowLens generates a short explanation of what the change does functionally and what it puts at risk: missing error handling, a removed retry, a credential that's no longer referenced correctly.

Missing error handlingRemoved retryCredential reference

At risk: failed charges now reach customers without a retry.

It proposes a fix, and you decide whether to apply it.

04

For common failure patterns, FlowLens suggests a specific repair, not a vague "check your configuration," an actual proposed change you can review before anything touches your live workflow.

Restore retry on HTTP Request 3Live untouched
Point it at the workflow that broke last

Free to start · No credit card

Your data, your model

Bring your own key.
Run it your way.

FlowLens doesn't require you to route your workflow data through a shared AI service you don't control. You can point it at your own OpenAI, Anthropic, or any OpenAI-compatible endpoint: your key, your usage, your cost. If you'd rather not send workflow data to a third-party model provider at all, the self-hosted path exists for exactly that reason.

This isn't a compliance checkbox. It's because workflow JSON often contains the shape of your actual business logic: which systems talk to which, what triggers what, and that's not something every team wants living in someone else's analytics pipeline by default.

AI providerSettings › Model

OpenAI

Your API key

Anthropic

Your API key

OpenAI-compatible

Any endpoint you choose

Self-hosted model

Never leaves your infra

https://api.openai.com/v1
sk-••••••••••••••••

Your key, your usage, your cost. Nothing requires a shared, FlowLens-operated AI service.

Integrations

Works with the platformsyou're already using

FlowLens connects to n8n, Zapier, and Make. Import a workflow by pasting its JSON, connecting directly, or syncing automatically, no migration, no rebuilding anything in a new tool.

n8ncloud · self-hosted
ZapierZaps
MakeScenarios

Paste its JSON

Drop an export in

Connect directly

Link your account or instance

Sync automatically

Every change becomes a snapshot

FAQ

Questions,
answered straight.

Something not covered here? Reach out rather than guessing from the docs.

Usually because something changed: a node parameter, a connection, a credential reference, and the change didn't surface anywhere obvious. FlowLens compares your current workflow against its last known-good version and shows you exactly what's different, instead of leaving you to scroll through the whole JSON tree manually.

See what changed in your last broken workflow.

Import it, compare it with its last good version, and read what actually changed.

Start free, no credit cardRead the self-hosted docs