Sigma Computing vs Countly

Sigma shipped a plugin for Claude Code that builds complete data models — metrics, relationships, columns, descriptions — from the terminal, alongside guidance on building Sigma Agents that handle schema discovery and model creation against Snowflake semantic views. Automated Actions landed for running reports, refreshing data, calling APIs, and triggering agents on a schedule, and embedded analytics gained bidirectional JavaScript events over postMessage.

Two directions are converging on the same idea: Sigma as a system that runs without someone watching it. Automated Actions handles the scheduled half, the Claude Code plugin and agent guidance handle the authored half, and the embedding work makes Sigma a component inside someone else's application rather than a destination. The recurring argument in the writing — that read-only dashboards are no longer enough — is consistent across all three.

Expect the agent surface to extend from model creation into model maintenance, since schema drift is what makes hand-built models rot. The embedded and automation threads suggest write-back workflows will keep deepening.