// documentation
agentic-sage documentation
SAGE is a passive, read-only fleet judge for parallel AI coding sessions (Claude Code, Grok Build, and friends). It watches and answers. It never edits, spawns, or blocks.
Site: sage.muslewski.com · npm: agentic-sage
Start here
| Path | For |
|---|---|
| Getting started | Install → sage init → sage on → sage doctor |
| Concepts: fleet judge | Why SAGE exists; default-OFF; human at fleet altitude |
| CLI reference | Verbs and flags |
| Configuration | Scope, storage, env |
| Adapters | Optional project enrichment |
| Recipes | Multi-harness, dogfood, statusline |
| Works with | Fleet siblings (herald, oracle, atlas, armory, ferry) |
Doctrine (short)
- Keep the human at fleet altitude — not a queen-agent orchestrator.
- Read-only / fail-open — missing data never blocks work.
- Default OFF — you opt sessions in; install does not hijack the machine.
- Advisory, not arbitrating — territory and merge briefs inform; they do not merge for you.
Where other knowledge lives
| Kind | Location |
|---|---|
| Public product docs | docs/ (this tree) |
| Architecture mind (Atlas) | agentic-sage-mind/ — zones, decisions, specs, plans |
| Agent install runbook | AGENTS.md |
| Human setup deep-dive | SETUP.md |
| Changelog | CHANGELOG.md |
Specs and implementation plans used to live under docs/superpowers/. They now live in the mind vault (Syndcast-style): see SUPERPOWERS-MOVED.md.