Error tracking
that answers back.
Atflow turns a flood of errors into one report each, then runs the flow you drew: tell the right person, retry what can be retried, update your status page.
- 1 flow free forever
- No card needed
- Live in ten minutes
1,204 renewals processed
No reports, nobody paged
A normal renewal: the flow runs end to end and stays quiet.
What you get
Everything between an error and its fix
Without writing your own alerting, retries and dashboards around every service.
One bad deploy and the alert channel is unreadable?
One report per bug, not a thousand
Identical failures collapse into a single report with a count, so a bug that fires 318 times is one line you can act on.
- Grouped by what failed, not when
- Open, reopened and resolved at a glance
- Every occurrence kept for the detail
payment-intent-failed
Billing · 2 min ago
webhook-timeout
Billing · 18 min ago
pdf-render-error
Invoices · 1 h ago
login-rate-limited
Auth · Yesterday
Your incident process lives in a wiki nobody reads?
Draw it once. It runs every time.
A flow is a picture of how your system works and what should happen when part of it fails — and it is the thing that actually runs.
- Drag-and-drop builder
- Each error lands on the node it came from
- Undo, redo and safe concurrent edits
The function throws. Every identical failure collapses into one unique report, then escalates once — to on-call and the status page.
The wrong people get woken up?
The right person hears, the right way
Choose per flow who hears about what: an email for every new report, a text for the critical ones, a Jira issue for the backlog.
- Email and SMS
- Jira issues and outgoing webhooks
- Escalates once, not per occurrence
Every new and reopened report
SMS
Critical issues only
Jira
Open an issue in BILLING
Webhook
POST to your incident tool
Too many reports to read them all?
Ask AI what to fix first
Ask in plain words what broke this week and what matters most. The answer comes from your own reports, not from guesses.
- Answers from your team’s reports only
- Secrets are stripped before anything is sent
- A weekly summary by email if you want one
What should we fix first this week?
payment-intent-failed — 318 times since Monday, all from the Billing flow's Lambda. It is still open.
Based on 4 open reports in your team
Your AI writes the code but knows nothing about how it fails?
Let your AI agents build the flow
Connect Claude Code, Cursor or any MCP client. It reads your flows and adds the steps for the service it just wrote.
- MCP server and REST API
- Read-only or read & write keys, with expiry
- From the Starter plan
> Add an "Invoice PDF" Lambda after the payments queue in my billing flow ● atflow · get_flow billing ● atflow · add_nodes +1 node "Invoice PDF" ● atflow · connect_nodes payments-queue → invoice-pdf Done. The billing flow has the new step, and its errors will be reported under it.
Support gets flooded every time something is down?
A status page your users can read
Show what is degraded and what is fine, on your own domain, so customers stop opening tickets to tell you what you already know.
- Uptime history per component
- Your own domain
- From the Starter plan
45 days agoToday
How it works
Three steps, about ten minutes
Nothing to install: your app sends errors to a webhook, and the flow does the rest.
Draw your flow
Add the parts of your system — APIs, queues, functions, databases — and connect them.
Point your app at it
Send errors to the flow’s webhook with one HTTP call. Any language, no SDK needed.
Hear about what matters
Reports are grouped and routed as you decided. You fix things instead of finding them.
Client feedback
What people say about it
Atflow is built by WNC. Two of these are clients; the third builds the thing.
“Our alerting used to be a firehose — one bad deploy and the channel was unusable, so people muted it. Atflow groups the identical failures into a single report, and the flow decides whether it is worth waking someone. We went from ignoring alerts to trusting them.”
Stephen Murphy
Founder, Arctik AI · 🇦🇪 Dubai, UAE
“What sold me was drawing the incident process instead of describing it in a runbook nobody reads. The flow is the documentation, and it is also the thing that runs. Retries, escalation, the status page — all in one picture I can hand to a new engineer.”
“We built Atflow because every project we shipped ended up with the same hand-rolled pile of error handling, retries and Slack webhooks. It runs on our own products first — if a flow is not good enough for our on-call, it does not ship.”
Pricing
Start free. Pay when it earns its place.
A base price, then pay only for the flows and new reports you add. Paid plans are never cut off.
Hobby
For trying Atflow
$0/ month
Start free- Flows
- 1
- Unique reports
- 40 / mo
- Team members
- 2
- History
- 30 days
Includes
- Visual flow builder
- Real-time reports
- Email & SMS alerts
- Jira & webhooks
- Ask AI about your issues
Starter
For small teams
$49/ month
Choose Starter- Flows
- 4+$10 each
- Unique reports
- 200 / mo+$0.20 each
- Team members
- 10
- History
- 90 days
Everything in Hobby, plus
- Public status page
- API & MCP server
- CSV export
- Compact view
- Support chat
Pro
For teams with on-call
$89/ month
Choose Pro- Flows
- 8+$8 each
- Unique reports
- 500 / mo+$0.15 each
- Team members
- 25
- History
- 180 days
Everything in Starter, plus
- Advanced analytics
- Uptime & cron monitors
- On-call & escalation
- Status page subscribers
- Private status pages
Enterprise
For compliance needs
from$289/ month
Talk to sales- Flows
- Unlimited
- Unique reports
- Committed
- Team members
- Unlimited
- History
- Custom
Everything in Pro, plus
- Audit logs
- Custom roles & SCIM
- EU data residency
- Uptime SLA
- Success manager
- On-premise option
Repeat occurrences of a report are always free. Paid plans are never cut off — you set a monthly cap on report overage (default $50).
Atflow is on your desktop too. Native apps for macOS, Windows and Linux bring reports to your system tray, on the same account.
FAQ
Questions people ask
- What is a Flow?
- A Flow is a visual workflow that defines how errors and events are processed in your system. Each flow can include multiple nodes for triggering, processing, and responding to events.
- What counts as a Unique Report?
- A Unique Report is a distinct error or event type. Multiple occurrences of the same error count as one unique report and are always free, so a bug that fires ten thousand times costs the same as one that fires once.
- What happens if I go over what my plan includes?
- On paid plans nothing stops: extra unique reports and extra flows are added to your bill at your plan’s price. You set a monthly cap on report overage (default $50), and a burst of new reports from one flow is grouped instead of billed one by one. On Hobby, reports past 40 are still kept, just locked until you upgrade.
- Can I upgrade or downgrade my plan?
- Yes, you can change your plan at any time. When upgrading, you get immediate access to new features. When downgrading, changes take effect at the end of your billing cycle.
- Do you offer refunds?
- The Hobby plan is free, so you can try Atflow before committing. For paid plans, we don't offer refunds, but you can cancel anytime and continue using the service until the end of your billing period.
- What integrations do you support?
- Email and SMS notifications, Jira issues, and outgoing webhooks to any HTTP endpoint. From Starter, AI agents such as Claude Code and Cursor can read and build your flows through the REST API and the MCP server.
- Is my data secure?
- Every connection is encrypted in transit, every team only ever sees its own flows and reports, integration credentials are encrypted at rest with AES-256, and API keys are stored only as hashes and can be scoped read-only, set to expire and revoked at any time. The details are in our Privacy Policy.
Start tracking in ten minutes
One flow free, no card required. Draw your first one and point your app at it.
Start free