v5.0 Documentation

ARCHIVED

You're viewing the archived v5.0 documentation. Switch to v5.1 (latest) for current features including codebase awareness, cross-project recall, and unified storage.

Installation

curl -fsSL https://quint.codes/install.sh | bash

# Then in your project
quint-code init            # Claude Code (default)
quint-code init --cursor   # Cursor
quint-code init --gemini   # Gemini CLI
quint-code init --codex    # Codex CLI

Decision Modes

/q-reason — describe your problem, the agent auto-selects depth (tactical, standard, or deep) and runs the decision cycle.

Full cycle: /q-frame → /q-explore → /q-compare → /q-decide

Quick cycle: /q-frame → /q-explore → /q-decide

Micro-note: The agent captures decisions automatically when it notices them. No rationale → no record. Auto-expires in 90 days.

Commands

Command	What it does
/q-reason	Start structured reasoning (auto-depth)
/q-note	Record micro-decision with rationale
/q-frame	Frame the problem
/q-explore	Generate distinct variants
/q-compare	Fair comparison on Pareto front
/q-decide	Record decision contract
/q-refresh	Manage artifact lifecycle
/q-status	Dashboard
/q-search	Full-text search

Key Concepts

• R_eff — computed trust score = min(evidence scores) with CL penalties. Expired evidence → 0.1.
• Evidence Decay — R_eff < 0.5 = stale. R_eff < 0.3 = AT RISK.
• Indicator Roles — constraint (hard limit), target (optimize), observation (Anti-Goodhart).
• Parity — same inputs, scope, budget for all options.
• DRR — four-component decision record: Problem Frame, Decision, Rationale, Consequences.

Storage (v5.0)

In v5.0, the database lives at .quint/quint.db inside your project. Markdown projections are in .quint/decisions/, .quint/problems/, etc.