Skip to content

The Observatory

Copy page

The Observatory is the console where you watch coordination between you and your agents, across projects and over time. It lives at /observatory, and it shows how a project’s shape is changing, what the observer sees, where coordination friction appears, whether memory is helping or failing, and which way things are trending.

It answers trajectory questions — is coordination getting smoother or rougher, is anything stabilizing into a pattern worth keeping — rather than point-in-time status like “is the API up right now,” which belongs in an operational dashboard.

The console models telemetry as variables over time. You pick a variable to Show, optionally overlay a second to compare it against, and choose how to view the result:

ControlWhat it sets
ShowThe variable to track
overlaid withA second variable to compare, or none
asTrend (a line over time) or relationship (a scatter, one point per day)

For whatever you select, an interpretation panel states what you’re looking at, why the variables relate, what patterns the view can expose, the observer’s current reading, and which signals aren’t instrumented yet.

VariableState
Architecture changesinstrumented
Working episodesinstrumented
Tool activityinstrumented
Observer candidatesinstrumented
Friction recurrencenot instrumented
Memory missesnot instrumented
Correction frequencynot instrumented

A variable that isn’t instrumented is still selectable, but has no data — the chart marks it not instrumented.

You look at the same patterns through different lenses — memory, architecture, coordination, observer, cross-project — each foregrounding one dimension and hiding the rest, so the thing you’re studying isn’t drowned out.

Open it and pick the variable that matches your question. The Observatory is your own deployment; the public demo runs on sample data as a read-only reference. To open it, run it locally, or feed it live data, see Run the Observatory and The Observatory feed.