I don't do the work.
A passive, read-only judge for fleets of parallel AI coding sessions. One judge per repo.
Run a fleet of coding agents at once and they collide — same files, duplicated work, lost context across sessions, no shared source of truth. Nobody is watching the fleet as a whole.
SAGE sits beside your fleet and watches every session. Ask it anything — who's touching what, where two branches diverged, what the backlog says — and it answers.
It never edits your code, never spawns agents, never blocks an action. Passive by design.
Once configured, you already benefit from SAGE — every coding session can orient itself around the others through the CLI. That’s layer 1. Layer 2 is optional: start a passive judge pane in the background, and those same CLI answers become a real AI judge.
Deterministic facts only — who is live, who claims which paths, why branches diverged. Sessions use board, territory, merge-brief and orient around each other. No extra pane required.
Start a passive agent session that watches the fleet and publishes short briefs into the same store. Workers keep the same CLI verbs — but answers gain narrative and per-audience advice. That’s what makes the CLI an actual AI judge.
Watches and answers. Never edits, spawns, or blocks. Your fleet keeps moving; SAGE keeps score.
It cannot change your repo. Ever. No hooks that write, no side effects, no surprises.
One Node CLI, no install bloat. Clone and run. Your package.json stays untouched.
Installing changes nothing until sage on. Try it without committing to it.
Any error in the optional guard allows the action through. SAGE is never in your way — even broken.
When idle the hook short-circuits instantly. Zero cost on the commits that matter most.
the sage sees all · the sage judges all
MIT · zero dependencies · one judge per repo
SAGE is in beta — early, active development. It works (I run it daily), but expect rough edges, and expect things to change as it finds its shape.
I'm building it in the open, and I'd genuinely love to hear from you. A bug, a question, a half-formed idea, or just hello — any form of contact is welcome. Nothing is too small.