Sopify · AI coding

Version AI coding plans and evidence in Git

Resume work across tools, teammates, and agents. Plans, decisions and verification evidence are kept as ordinary files under .sopify/ and versioned in Git, so the next session resumes from the record instead of from memory.

Install Sopify View on GitHub

Installs into the AI coding tools you already use. Questions and small fixes are handled directly; plans start only when needed.

Hand-drawn Sopify workspace showing an AI workflow that pauses, records a plan, and resumes
plan → checkpoint → handoff → resume

What it prevents

Move fast without avoidable rework

Sopify keeps fast AI coding from outrunning the facts, the decision, or the project history.

01 / ASK

Premature coding

When a requirement or key choice is unresolved, Sopify pauses instead of guessing.

02 / RESUME

Lost context

Plans and evidence are versioned in Git, so a new session, tool, machine, or teammate can resume from the same record.

03 / TRACE

Forgotten decisions

Requirements, tradeoffs, reviews, and verification evidence remain inspectable project assets.

How it works

A workflow with deliberate stopping points

Simple requests stay simple. Work that needs a plan moves through four explicit stages and pauses when you need to decide.

STEP 01

Analyze

Clarify the outcome, evidence, and unresolved boundaries.

STEP 02

Design

Choose the smallest sufficient path and make decisions visible.

STEP 03

Develop

Implement the approved tasks and attach focused verification.

STEP 04

Finalize

Close the plan only when delivery evidence is ready.

Sopify workflow showing direct requests, managed plans, checkpoints, and finalization

Product form

The host executes. Sopify keeps the work resumable

Sopify is a set of rules and Skills that installs into the AI coding tool you already use. It preserves plans, decisions, and verification evidence as project files in .sopify/.

~go runs or resumes the current plan from those files; routine questions and small fixes are answered directly without resuming a plan. A local handoff pointer provides a same-workspace resume hint.

.sopify/plan/…
  • Codex · verified
  • Claude Code · verified
  • Qoder · verified
  • Copilot · baseline
  • Cursor · baseline
  • Antigravity · baseline
  • Trae · baseline
Sopify architecture showing host adapters, shared workflow skills, writer boundaries, and repository state Open the technical architecture →

Install

Add Sopify to the host you already use

The installer checks for Python 3.11 or newer before downloading. You can inspect the script before running it.

curl -fsSL https://github.com/evidentloop/sopify/releases/latest/download/install.sh | bash -s -- --target codex:en-US

FAQ

The short version

How does Sopify work with existing AI coding tools?

Sopify is a set of rules and Skills that installs into the AI coding tools you already use.

Does every question start a plan?

No. Ordinary questions and small fixes are handled directly. Sopify resumes an unfinished plan only when you ask to continue or use the current host's managed entry.

What actually travels with the repo?

Plans, decisions, and verification receipts can be tracked under .sopify/. The active-plan and handoff pointers stay gitignored by design.

Does switching hosts give every host identical capabilities?

No. Codex, Claude Code and Qoder are verified: the managed workflow has been run in a real session, not only installed. Copilot, Cursor, Antigravity and Trae are baseline: the installation is verified, and real-session evidence varies by host — see each host doc; Antigravity has been tested in real sessions. One Cursor install covers the IDE and local CLI; the CLI selects the plan, development, resume, and closeout entry when the request calls for it. Cloud Agent remains out of scope.

Is EvidentLoop required?

No. Sopify works on its own. EvidentLoop is an optional companion for auditable reviews of local Git changes.