Automations — your agent, on a schedule
An automation is a prompt you hand over once, with a schedule attached. After that, the same agent runs it on its own — and every run is saved as a conversation you can open and read.
What are automations?
An automation runs the same agent you chat with — same skills, same tools, same ability to search the web and write — just unattended. You describe the job once; from then on it fires on its schedule and leaves the results where you can read them.
What that looks like in practice:
- ☀️ A daily morning brief — the news you care about, summarized before you sit down.
- 📋 A weekly report — research refreshed and rewritten every Monday at 9.
- ⏰ A one-time reminder — something tonight at 8, once, then it's done.
Every run is saved as a normal conversation — what the agent searched, what it wrote, what it produced. Nothing runs in a black box.
Two ways to create one
1. Just say it, in chat
The fastest way is to ask, in plain language, in any conversation:
"Every Monday at 9, give me a weekly report on my industry — competitors, funding, product launches, in 10 bullets."
The agent creates the automation for you and confirms exactly what it scheduled: the prompt, the cadence, the time. If that's not what you meant, say so and it adjusts.
2. The Automations panel
Prefer to point and click? Inside the app, open the Automations panel on the left side and press +. Write the prompt, pick a schedule, save. Same result as asking in chat.
Schedule options
Four cadences cover most recurring work:
At a specific time — tonight at 8:00 PM, or next Tuesday at 9:00 AM.
For monitors and watchlists — check every hour and only speak up when something changed.
At HH:MM you choose — the classic morning brief.
On a chosen day at HH:MM — reports that land when your week starts.
Schedules run in your timezone — the one your browser reports. Set a run for 9:00 and it fires at 9:00 where you are, without conversion math.
Run history
Every run is saved as a conversation, listed under its automation. Click one and read it like any other: which sources the agent pulled, what it wrote, which files it produced. Skim the latest run at breakfast; dig into last week's when you need to.
If a run fails, it's marked in red — you see that it failed and why, not a silent gap in the list.
Pause, resume, delete
Going on vacation? Pause the automation and it stops firing. When you're back, resume it and the next scheduled run picks up from there. Done with a task for good? Delete it — permanent, and immediate.
Credits
Every run consumes credits and is billed exactly like chat — a run that researches the web and writes a full report costs about what the same request would cost if you typed it yourself. A short daily brief costs little; a deep weekly research report costs more. No separate subscription, no per-automation fee.
Set one up in under a minute
Open the app, tell the agent what you want and when you want it — it takes care of the rest.
Open the app