An analyst that reads everything, and asks before it acts.
Atlas is the agent inside Patena. It reads your organization’s stream, answers with citations, runs on a schedule for every analyst, and proposes changes instead of making them. The dashboard is where you check its work, not where the work happens.
Three things changed. The dockworker strike reached Gdańsk and Riga this morning 1; two of your suppliers ship through Gdańsk 2. The EU sanctions annex now lists Halvard Shipping AS, a carrier both of them use 3. Tallinn remains open; reports of a stoppage there come from one T3 source and I am not counting them 4.
One decision: the “Baltic ports” monitor has no threshold for strikes. I have drafted one (Unrest · severity ≥ 3 · Gdańsk, Riga, Tallinn) and will add it only if you approve.
Atlas answering a morning question for a supply-chain analyst: three cited changes, one unconfirmed report flagged as such, one proposed monitor waiting for approval. The right panel is the trace of every tool call it made.
Three ways an agent works for you
The same Atlas, in three postures. All of them read only what the owning analyst can read.
Ask
A question in plain language, in any of the thirteen languages. Atlas searches stories, entities, scores and your monitors, then answers in text with a citation on every claim. No source, no sentence.
Watch
A standing instruction with a schedule. Every 15 minutes, hourly or daily, the agent re-reads the window, compares it with your thresholds and your last acknowledgement, and notifies only on change.
Draft
A report, a brief, a monitor or a subject, prepared and left for review. Agents write; people publish. Nothing leaves the organization and nothing in its configuration changes without a human click.
Your agents, this week
A per-user roster. Each analyst sees their own agents; org admins see all of them, with the audit log.
What an agent may and may not do
The rules are the same for every agent on the platform and cannot be loosened per organization.
An agent sees exactly the stories, entities, scores and private sources its owner can see. Seats and roles apply to agents as to people.
Every sentence in an answer carries a source. If the stream does not support a claim, Atlas says it does not know, with the count it did find.
Monitors, subjects, thresholds, delivery and sources change only when a person approves. Proposals expire in 7 days.
Each run is logged: tool calls, sources read, time, cost. The org audit log shows who approved what, and when.
Put an analyst on the stream.
Early access by request. We set up your organization and the first agents with you.