Integration

n8n

Know when your n8n workflow silently stops.

Import examples/n8n/ned-register-watches.json, set your callback URL in both nodes, run it once.
Import examples/n8n/ned-watched-workflow.json, paste the two watch ids, add two Header Auth credentials.
Move your steps in between "Ned: run started" and "Ned: run finished".

Both files use only core nodes (Schedule Trigger, Manual Trigger, No Op, HTTP Request), so no community nodes are needed.

The watched workflow

Schedule Trigger -> Ned: run started -> [your steps] -> Ned: run finished -> Ned: check in

node	call	what it means
Ned: run started	`POST /v1/watches/<overrun id>/start` with `{"run_id": "n8n-{{ $execution.id }}"}`	overrun: Ned fires if "run finished" doesn't follow within the limit
Ned: run finished	`POST /v1/watches/<overrun id>/finish` with the same run id	closes the run (a late finish sends `clear` with the runtime)
Ned: check in	`POST /v1/checkin/<deadman id>`	deadman: Ned fires if this doesn't arrive within the interval

"Ned: check in" is last, so it only runs when every step before it succeeded. A failing workflow therefore never checks in, and Ned tells you once the interval passes. (To report a failure at once, an Error Trigger workflow can POST {"run_id": "n8n-<execution id>", "status": "failed", "error": "..."} to the overrun's finish URL.)

Every Ned node is set to retry 3 times and On Error: Continue, so a problem reaching Ned never fails your workflow. Each sends X-Ned-Ref: n8n.

Setup, step by step

Import ned-register-watches.json. In both HTTP nodes replace https://your-agent.example/hooks/ned with your

callback URL, and my-n8n-workflow with a name for the job. Run it once.

"Register deadman" answers with watch_id, signing_secret and agent_key. Keep all three.
"Register overrun" uses that agent_key, so both watches belong to one agent. Keep its watch_id and signing_secret.

Create two credentials of type Header Auth: name Ned deadman secret (Name Authorization, Value

Bearer <deadman signing_secret>) and Ned overrun secret (the same with the overrun's secret). The secrets live in n8n's credential store, not in the workflow.

Import ned-watched-workflow.json. Replace w_OVERRUN_ID and w_DEADMAN_ID in the URLs, pick the two credentials

in the Ned nodes, set the schedule, and put your own steps where "Your work goes here" is.

The deadman is registered with "arm": true, so its clock started at registration: a workflow that never runs is

noticed too.

Only need one of the two? Delete the other nodes: a deadman alone is just the last node.

Notes

Interval: the deadman's interval_s (3600 in the template) should be a bit longer than your schedule.
EU: use https://api-eu.ned.watch in every URL.
The JSON was checked against parameter names taken from n8n's own node sources (HTTP Request v4.2, Schedule Trigger v1.2), and each HTTP call was replayed against a local Ned Watch. It hasn't been imported into a running n8n yet.