It lives, even when
no one is talking to it.
Pulsar is the autonomy layer: a background loop that runs watches, monitoring, and missions — with no human intervention.
Collect, compare, decide
No blind checks at fixed times. On each beat, Pulsar fetches fresh data, compares it to what it observed last tick, and only acts if a real change is detected. Nothing new? Radio silence — no needless noise.
Three trajectories
Continuous monitoring
Configurable frequency. Runs indefinitely on the designated target.
"Monitor this blog and summarize as soon as an article is published."
Until a goal
Once reached, the agent acts one last time and stops.
"Alert me when this product drops below €50, then stop."
Defined duration
Auto-stop. Suited for campaigns or launches.
"During the sale, monitor these 10 products and alert on each promo."
A long-term goal, pursued alone
Beyond monitoring, you can give Pumaï a heading: a goal, the allowed resources and channels, the constraints. It then advances on its own, tick after tick.
You scope the mission: goal, resources, allowed and forbidden channels, guardrails.
On each tick, its observations and actions serve the goal. It relies only on the listed resources — no inventing a source.
It journals each step forward — which feeds the next tick's context.
Any sensitive action is proposed for approval, with its rationale. You stay in control.
An autonomous watch does not read the assistant's private memory — and vice-versa: each profile keeps its own. What they share on purpose is the document cabinet and the company directory, to stay aligned on the same files. And when a watch fires, a summary lands in the assistant's inbox, never a dump of raw data.