00:00 - 04:00 - THE MIDWATCH - SOMEONE STAYS AWAKE
Your automations fail silently. Midwatch doesn't sleep.
Read-only monitoring for n8n. Silent stops, zombie successes, config drift - caught in minutes, not days. Connect an API key and the first check runs within ten minutes. Nothing installed in your workflows.
● CRITICAL Acme - Missed Call Textback - silent for 3h 42m (expected hourly)
▲ DRIFT Invoice Chaser - definition changed, no blessed baseline
✓ OK 18 workflows on watch - last cycle 2m ago
Read-only by design. Midwatch authenticates with a read-only key, stores workflow metadata and execution rollups - never your execution payloads, never your data. It cannot edit, execute, or delete anything.
How we handle your fleet ->The failures your Error Trigger can't see
Silent stops
Schedules that stop firing - on n8n or Cloudflare Workers - with no error to trigger on. No execution, no exception, nothing in the log.
Zombie successes
Green checkmark, zero output. The workflow 'succeeds' while processing nothing - AI agent steps are the worst offenders.
Config drift
Someone edits a node on a Friday afternoon. Nothing tells you what changed.
Policy violations
A workflow quietly references credentials it should never touch.
n8n's error workflows only fire when something errors. Most fleet damage never does.
On watch in five minutes
01
Connect a read-only API key - five minutes, nothing installed in your workflows.
02
Five checks, every five minutes - hard failures, silence, zombies, drift, reachability - from the Cloudflare edge, outside your instance, where your webhooks originate.
03
Alerts only when it matters - severity ladder, dedup and cooldown, SMS (Agency) for revenue-critical, a daily digest as the heartbeat.
Messages
Midwatch
CRITICAL: 'Acme - Missed Call Textback' silent 3h 42m (expected hourly). https://app.midwatch.ai/alerts
Know in minutes, not at tomorrow's digest.
Your fleet gets checked every five minutes. Revenue-critical failures page you by SMS (Agency); everything else lands in a deduplicated email digest that is itself the dead-man signal - if the digest ever goes missing, something is wrong and you know that too.
Dogfooding on our own agency fleet cut worst-case detection from 13 hours to under 10 minutes.
Someone changed a workflow. Now you know.
Every definition is hashed and compared against the baseline you blessed. A change - any change - raises a drift alert with a diff. Policy rules go further: declare what must never happen ('only TrueVig workflows may reference the TrueVig database') and Midwatch pages you the moment a definition violates it. Baselines are yours to bless; Midwatch never auto-accepts anything.
Invoice Chaser - drift detected 14:32
- "url": "https://api.client-crm.com/v2/invoices"
+ "url": "https://api.client-crm.com/v2/invoices-test"
Baseline blessed 2026-06-12 by jason - Re-bless or investigate
RUNNING CLIENT AUTOMATIONS? THE AGENCY PLAN ADDS THIS
The monthly report that renews the retainer.
Every other monitoring tool sells you an inward-facing dashboard. The Agency plan also writes the outward-facing artifact: a white-label monthly report for each client proving the automations ran, saved N hours, and were fixed before anyone noticed. Your name on it, your clients reading it, your retainer renewing.
See the sample reportMonthly Automation Report
Lakeside HVAC - June 2026
41.5 hrs
saved
1,284
executions
2
incidents caught in < 10 min
Prepared by Midwatch
<=10 min
detection, was 13 hrs
5 min
check cycle
0 writes
read-only, enforced by the key
5
failure classes watched
0
execution payloads stored
Two plans. Both start free.
Solo
$49/mo
14-day free trial - no card, cancel anytime
- 1 n8n instance or Cloudflare account
- Every detection, nothing installed
- Email / Slack / Telegram alerts
- Connect a read-only key; watching in ten minutes
Agency
$197/mo
Same 14-day free trial. Built for fleets with clients attached.
- 10 instances (n8n or Cloudflare accounts)
- Multi-client scoping
- White-label client ROI reports + SMS alerts
- Policy rules + multi-client alert routing
Questions, answered plainly
Is Midwatch really read-only?
Yes. Midwatch authenticates with an n8n API key and only ever issues GET requests (read-only by token scope for Cloudflare). It cannot edit, execute, enable, disable, or delete anything in your instance. The read-only guarantee is enforced in code, not policy - there is no write path to enforce it against. Details on the Security page.
What data do you store?
Workflow metadata (names, ids, active state, definition hashes), execution rollups (counts, timestamps, statuses), and alert history. Never execution payload bodies, never the data flowing through your workflows, never credential values.
Does it work with n8n Cloud and self-hosted?
Both. If your n8n exposes the REST API and can issue an API key, Midwatch can watch it - self-hosted, n8n Cloud, or a mix across clients.
How is this different from n8n error workflows?
An Error Trigger fires when a workflow errors. Most fleet damage never errors: a schedule silently stops, a "successful" run processes zero items, someone edits a node and nothing announces it. Midwatch watches for silence, zombies, drift, and policy violations from outside the instance - and adds the client-facing monthly report no error workflow can produce.
What happens when Midwatch finds drift?
You get an alert with a diff against the last baseline you blessed. You either re-bless the new state (it becomes the baseline) or go fix it. Midwatch never changes anything itself - it tells you, you decide.
I'm not an agency - is Midwatch for me?
Yes. The Solo plan exists for exactly you: one n8n instance or Cloudflare account, your own automations, every detection, alerts to your email, Slack, or Telegram. If your workflows book appointments, send invoices, or feed your CRM, a silent failure costs you money whether or not a client is watching. Connect a read-only key and the first check runs within ten minutes.
What is the guided setup?
An optional path for agencies: a short call with the founder to map your fleet, connect your instances together, and set up client scoping and reports - instead of doing it self-serve. Same plans, same prices. Request it at /apply, or skip the call entirely and start a 14-day free trial with no card.
What will it cost after the free trial?
Solo is $49/mo (1 n8n instance or Cloudflare account, every detection, email/Slack/Telegram alerts). Agency is $197/mo (10 instances (n8n or Cloudflare accounts), multi-client scoping, white-label client ROI reports, SMS alerts, policy rules). The self-serve trial is 14 days and does not require a card. Charges appear on your card statement as REVENUEWITHAI.COM.
Who is behind Midwatch?
Midwatch is a Revenue With AI service. We run n8n automations for our own clients, got burned by silent failures, and built the watch we could not buy. It monitored our live fleet for two months before self-serve signup opened.
Put a watch on your automations.
14-day free trial, no card. Read-only by design. Connect a key; the watch starts within ten minutes.
Start free trial