Manual
Fire it yourself
Global hotkeys and a command palette for the work you want to stay in control of. One keystroke away.
run flow…
- Invoice triage⌥⌘I
- Weekly recap⌘⇧R
- Clipboard → task⌥⌘T
- Screenshot to notes⌥⌘S
Desktop automation
Build automations and agents on your desktop.
Write down what should happen. AutomationGhost builds the steps on your machine, and you decide how they run: a hotkey, a schedule, or an agent that picks up the work while you are away.
01 · Plan first
You see the plan before anything runs.
Each request becomes a short written plan: the steps, the code it wants to write, and the tools it will use. Approve it, then the flow gets built.
You pull the trigger: hotkey, palette, or a pinned button.
Your plan
weekly-recap.plan01
Phase 01·Kick off
trigger.manual
You decide when it runs. No schedule, no webhook, just you
01aHotkey
hotkey.modifier+I
Press ⌘⇧I from anywhere in the desktop app
01bCommand palette
palette.run_flow
Search your flows and hit Enter to launch
01cToolbar pin
ui.button.run
One click on a pinned action in the toolbar
02
Phase 02·Confirm
prompt.confirm
Quick prompt for week range or recipient before it runs
03
Phase 03·Run
pipeline.execute
Snapshot metrics, format the recap, and stage the output
04
Phase 04·Deliver
action.copy_and_notify
Copy to clipboard and drop a Slack summary when done
The flow
7 nodes · 6 edgescompiled.flowready · manual
Automated
Let the clock run it
Schedules and events keep the routine going while you are away from the keyboard.
next run · weekdays 09:00
Agent
Hand it off
An agent watches the run, picks the right tools, and leaves you a report you can actually read.
warnings · 3
errors · 2
03 · Building blocks
Add the pieces your automation needs.
Need Slack, a PDF parser, a CRM lookup, or an agent that researches a lead? Drag those in as steps. Your flow stays one board you can open and change later.
invoice-tracker.flow · 5 nodes
04 · Outputs
Every run leaves something you can read.
Results show up as notes, stats, tables, and timelines. The Gallery in the app keeps the latest run per flow in one place, so you do not have to dig through logs.
3 / 3 flows · Gallery output
Invoice follow-up
New invoice email parsed. Amount and due date logged. Team notified if overdue.
invoice-triage · live · Gmail · new email
| Client | Status |
|---|---|
| Acme Studio | Paid |
| Beta Works | Overdue |
| Northline | Sent |
New invoice email parsed. Amount and due date logged. Team notified for Beta Works.
Today
Logged & notified
Wed
Reminder email
Fri
Escalate if needed
Same pin layouts as the desktop app Gallery: markdown, stat cards, tables, and timelines — one readable surface per flow run.
FAQ
Questions worth asking.
Do I need to know how to code?
No. You describe the automation in plain language. The app drafts a plan, then builds it. The result is TypeScript on disk, so if you do read code you can open any step and change it.
Where do my API keys live?
In your operating system keychain. They are read on your machine at run time and sent only to the provider you set up. They never pass through our servers, because there are no servers in the run path.
What happens if I cancel?
The desktop app keeps working for local and manual runs, and your flows stay on disk as files you own. You lose Relay event forwarding and Cloud sandboxes.
Can it reach my internal tools?
Yes. That is the point of running locally. Flows run inside your network, so anything your machine can reach (an intranet API, a VPN service, a local database) is reachable without opening it to the internet.
Which models can I use?
Any provider you hold a key for, plus local models through LM Studio. You can use one model for planning and another for the run steps to control cost.
How is this different from Zapier or n8n?
Those run everything on their servers and charge per task. AutomationGhost starts as a desktop app with real code steps on your machine. Relay only forwards outside events to you. Cloud is optional if you want managed sandboxes.