I made a standing routine something you describe in a sentence, not something you script.
How it was built, and why
Treft turns a plain sentence into a standing routine. Describe a job, no triggers to wire and nothing to script, and the platform drafts the model, the behaviour profile and a schedule you can edit. Every agent lives on one wall with status and next run visible.
The flow
Write the brief in plain language, the platform configures the model, behaviour and timing, a dry run checks it on live inputs before anything counts, then the finished work lands each morning in a single digest card.
What it is
An agent platform with a plain-language creation flow and a single wall for every agent.
Plain-language agent creation with a dry run before it counts
One wall showing status, last run and next run
A single morning digest backed by step-by-step run detail with timing and token counts
How it works
I kept the surface calm on purpose: one wall instead of a dashboard grid, one digest card instead of an inbox, and a run-detail view that plays back step by step when someone needs the full trace.
The creation flow drafts the configuration from the brief, so the hard part is describing the job well, not wiring the machinery.
Why it matters
Standing automations usually demand triggers, scripts and constant babysitting. Letting someone brief a job in a sentence and read the result in one card is what makes it something a non-engineer will actually keep using.
