Prove what broke, and whose fault it was.
Docket watches every automation you run for clients, catches the failures n8n marks green, and gives you a monthly statement per client showing what went wrong, who caused it, and whether it is billable.
Docket ships in about six weeks. Five founding customers are being taken now at €249 a month, held for as long as they stay.
Apply for a founding placeBuilt for n8n first. No other platform is promised yet.
The unpaid hour before you can even decide to bill
“my workflow bug I eat it. Client credential expiry I bill for.”
“if there is no monitoring, you usually only hear about it when the client complains, and that turns into reactive support very quickly.”
The rule is easy. Applying it is not. When something breaks, you go into the execution log and spend an hour working out whose fault it was, before you can even decide whether the fix is billable. That hour is never invoiced. Do it twice a month across a dozen retainers and it is a working week a year you gave away, on top of the fixes you swallowed because you could not prove they were not yours.
The fault ledger
One statement per client per month. Every failure, what it affected, what caused it, and whether it is yours to eat or theirs to pay for. It is a document you can send, not a screen you have to interpret.
Example document · illustrative data · fictional client
Automation fault ledger
Northgate Dental
1 August to 31 August 2026 · 6 workflows · 4,180 runs
Prepared by
Your agency name
Issued 01 September 2026
| When | What it affected | Cause | Billable |
|---|---|---|---|
| 03 Aug, 09:12 | Appointment reminders Twilio credential rejected. 41 reminders not sent. | Client credential | Yes |
| 07 Aug, 02:04 | Nightly patient sync Practice API changed the date format. 18 records written with an empty date of birth. | Upstream API | Yes |
| 12 Aug, 14:38 | New patient intake Run finished green and wrote nothing. Our filter excluded every record after a field rename on our side. | Our logic | No |
| 19 Aug, 06:55 | Invoice chase Sheets API quota exceeded. Recovered on the third retry, 41 minutes late. | Upstream API | No |
| 26 Aug, 11:20 | Appointment reminders Twilio key rotated on the client side without notice. 12 reminders not sent. | Client credential | Yes |
5 failures this period. 3 attributable to causes outside your workflows.
Billable remediation: 3 of 5
Northgate Dental is invented, and so is every figure above. Docket has no customers yet, so there is no real ledger to show you.
Margin per client
What each client costs you to run, set against what they pay you. Retainers are priced once and then never looked at again, which is how an account ends up costing more to serve than it brings in without anybody noticing.
Example · illustrative data · fictional clients
| Client | Retainer | Run + token cost | Margin |
|---|---|---|---|
| Northgate Dental | €450 | €38 | €412 |
| Ardmore Physio | €350 | €44 | €306 |
| Halloran Motors | €300 | €61 | €239 |
| Brightside Lettings | €250 | €117 | €133 |
| Carraig Foods | €180 | €214 | -€34 |
Cost here is infrastructure and model spend only. It is not your time, so a row that is barely positive is usually losing money once you count the hours.
How it works
Connect each client instance
One read-only n8n API key per client. Cloud or self-hosted, both work. Docket reads execution history. It does not edit your workflows and needs no inbound access to your servers.
Check outcomes, not exit codes
A run that n8n marks green but wrote nothing is a failure. So is one that returned the right shape with the wrong values. Docket checks what the run actually produced against what that workflow normally produces, so silent failures surface instead of sitting there for three weeks.
Attribute the cause
Every failure is classified as the client's credential, a change upstream, or your own logic. The execution record, the error, and the run before it are attached as evidence. That is the part that decides whether a fix is billable.
Bill it and report it
At month end each client gets a statement in your branding. You get the same failures priced, and margin across your whole book of retainers.
Pricing
Per agency, per month, whatever the number of runs. Prices exclude VAT, which is calculated at checkout based on your location.
Monitoring
The floor. Know that something broke.
€99/ month
- Failure detection across all client instances
- Catches runs n8n marks green that did nothing
- Alerts to email or Slack
- Full run history, searchable across clients
Ledger
RecommendedThe product. Know whose fault it was, and what it earned you.
€299/ month
- Everything in Monitoring
- Fault attribution: client credential, upstream API, or your own logic
- The monthly fault ledger, one statement per client
- White-label client reporting in your branding
- Margin per client against what they pay you
- Up to 15 clients
Agency
The same product, for a bigger book.
€499/ month
- Everything in Ledger
- Up to 50 clients
- Priority support
None of these are on sale yet. Docket is in development and the only way in today is a founding place.
Five founding customers
Docket ships in about six weeks. Five agencies are being taken on before then at €249 a month, and that price is held for as long as they stay subscribed.
In exchange: a short call every week while it is being built, and a named case study once it is working for you. You will be shaping what the ledger looks like, which is worth more to you than the discount.
Would rather just email? alan@devlar.io
Questions
When does it ship?
About six weeks from now. Founding customers get access to each piece as it lands rather than waiting for the whole thing, and the fault ledger is the first thing being built.
Does it work with self-hosted n8n?
Yes. Docket needs an n8n API key with read access to executions and workflows, and it calls your instance over HTTPS. There is no agent to install and no inbound firewall rule to open. Self-hosted and n8n Cloud are both supported. Two honest caveats while this is still being built. Catching a run that finished green means reading successful executions, so an instance configured to discard those will need that setting changed. And read-only access is a commitment that will hold, but if attribution for some workflow turns out to need something added on your side, founding customers will hear that from me before they hit it rather than after.
Does our clients' workflow or prompt content leave their infrastructure?
By default Docket stores execution metadata only: workflow and node names, timestamps, status, error messages, and token and run counts. Capturing field-level payloads is off by default and turned on per workflow when you want the evidence in the ledger to be specific. Data is held in the EU. This is the design being built to, and it is one of the things founding customers get to argue with before it is fixed.
We already get an email when a workflow fails. What does Docket add?
Knowing something failed does not tell you whose fault it was, and that is the part you cannot bill without. Docket does the attribution and puts it in a document you can send to the client. It also joins the cost of running those automations to what the client pays you, which nothing in your n8n instance can tell you.
What happens to founding pricing later?
€249 a month for as long as you stay subscribed. It does not step up after a year and it does not move when the list price moves. If you cancel and come back, you come back at list price.
What about Make, Zapier, or Power Automate?
n8n only for now, and there is no date for anything else. Attribution depends on reading execution history in detail, and doing that properly for one platform is worth more than doing it thinly for four.