n8n monitoring for agencies

Every client's n8n,
on one screen.

You built the automations, handed them over, and you are still the one who gets called when one goes quiet. AgentPing watches every client workflow from one console, pages you the moment a scheduled run goes missing, and gives you the uptime record to show for it.

Track your first client workflow How n8n monitoring works
support-triage · schedulelive
support-triage missed its 14:00 run paged on-call · last ok 13:00 · 247 runs clean before

Ten clients, forty workflows, and no shared view.

Multi-client automation work has a monitoring problem that single-team advice does not solve. The workflows live in instances you may not control, on schedules you set months ago, for clients who assume silence means everything is fine.

The instances are not yours

Some clients run n8n cloud, some self-host, some moved it in-house after handover. Checking each executions list means a login per client and a habit nobody keeps past the second month.

The failure you care about leaves no trace

A deactivated workflow produces no failed execution, because it produces no execution at all. There is nothing in the client's n8n to find, which is why these run for weeks.

You find out from the client

The worst version of this conversation starts with them telling you. By then the question is not whether the workflow broke, it is why you did not know, and that is the part that costs the retainer.

One team per client, one login for you.

Teams are unlimited on every AgentPing plan, including Free. A team per client keeps each roster of workflows separate while you keep a single console across all of them.

1. A team per client

Create a team named for the client. Their workflows, alert routes and status page live inside it, so nothing about one client is visible from another.

2. A ping token per workflow

Issue a per-agent ping token, not an API key. It is scoped to the one workflow, so it is safe sitting in a URL inside a client's n8n, and revoking it affects nothing else.

3. One HTTP Request node at the end of the workflow

Add the node before handover and the monitoring outlives your access. Give the agent its expected cadence in AgentPing and a run that never arrives raises an incident on its own.

example · http request node n8n 
# One node at the end of each client workflow.
curl "https://eu.ingest.agentping.io/v1/ping?key=ping_…&agent=acme-lead-enrichment&status=ok"

# Add the model and token counts and we price the run for you:
curl "https://eu.ingest.agentping.io/v1/ping?key=ping_…&agent=acme-lead-enrichment&status=ok\
&provider=openai&model=gpt-4o&input_tokens=1240&output_tokens=380"

Full setup is in the n8n docs, and the n8n monitoring page covers the failure modes in more detail.

Turn uptime into something you can show.

Every heartbeat is a dated record. Over a quarter that becomes the evidence behind the retainer: what should have run, what did, what was missed and how fast you caught it.

A status page per client

A link-shareable view of the workflows you choose, with 90 days of history. Passcode-protect it, regenerate the link to revoke it. States only, never payloads or costs.

AI cost per client

Spend attributed to the client that caused it, priced server side. A defensible number to pass through, or to check the retainer still covers the work.

Incidents with a paper trail

Each missed run opens an incident and closes when it recovers, so "we caught it in eleven minutes" is a record rather than a claim.

How do I monitor n8n workflows for multiple clients at once?
Create one team per client inside a single AgentPing account. Teams are unlimited on every plan, including Free, so a twenty-client roster costs the same as a two-client one. Each client's workflows report into their own team, and you keep one login that sees all of them.
Do I need access to the client's n8n instance?
No. The workflow reports outbound to AgentPing from an HTTP Request node, so it works whether the client runs n8n cloud or self-hosts, and whether or not you still have a login. That is the point: the monitoring survives the handover.
Is it safe to put a key in a client workflow?
Use a ping token rather than an API key. A ping token is scoped to one agent, so it is safe to sit in a URL inside a client's workflow, and if it leaks the worst case is a false heartbeat for that one workflow. It cannot read data or reach anything else in your account.
Can I give a client proof their automations are running?
Yes. A status page is a public, link-shareable view of the workflows you choose, showing operational, degraded or down with 90 days of history each. Add a passcode to keep it private and regenerate the link to revoke access. It shows states only, never payloads or costs, so one client never sees another. One page on Starter, five on Team, unlimited on Business.
Can I bill AI cost back to each client?
Yes. Send the provider, model and token counts on the ping and AgentPing prices the run server side from a maintained rate card. Spend rolls up per client without any cost arithmetic inside n8n, which gives you a defensible figure to pass through or price into the retainer.
What does it cost to monitor a lot of client workflows?
Agents are unlimited from the Team plan at $199/month, so the number of client workflows you watch does not change the bill. Starter at $99 covers 10 agents if you are just starting out, and the free plan covers 2 while you try it. Named limits, no metered billing, no per-seat charge.

Never hear it from the client first.

Add one client workflow, give it an expected cadence, and let AgentPing tell you the moment it misses. Free plan, no card required.

Track your first client workflow

n8n monitoring → For automation agencies → Explore Pulse →