Meerkats AI
Use cases

What teams build on Meerkats.

Systems built by customers on the same six layers. Names are withheld; the architecture is described layer by layer. Every number is computed by code, every action is approved, and every run can be opened.

What they share

Different systems. The same foundation.

One definition per metric and entity

The catalog is the single name authority. Both systems report numbers nobody has to reconcile.

Context loaded, not rebuilt

Each run reads the timeline, the state and the rules from the context layer. The model never sees raw rows.

Typed automations with state

Both are contracts with triggers, conditions, caps and approvals. Both remember what they did last run.

Actions from a registry

Every write is a registered, permissioned operation with a verify read and a rollback. Nothing else can write.

A person in the loop by policy

Low-risk, reversible actions run alone. Anything else stages a plan and waits.

A record of every run

Inputs with sync time, the rule, the plan, the approver, the change, the read-back, and the outcome afterwards.

Build the next one on the same six layers.

Connect a system, define a metric, declare an automation. The first run is traceable from the start.