尚未收集仓库信息。
尚未检测到更新。GitHub 发布版本后会显示在这里。
YallaFlow — Persistent project memory and verified development workflows for AI coding agents like Claude Code and Codex.
翻译准备中
正在加载完整 README。
查看完整 README ↗Your coding agent writes code fast. Then the next session forgets your project, and it says "done" without proof.
YallaFlow fixes that. It gives Claude Code and Codex a reviewed memory of your codebase, a workflow sized to each request, and evidence before anything is called done. A local console, in English and Arabic, lets you watch every request move and start your app with one button.
It runs locally, sends no telemetry, and keeps everything in your repository.
yallaflow console opens a local web app in Arabic or English, light or dark. It shows every project, every request and every running service on one screen. Nothing runs on its own, and nothing leaves your machine.
| Create work in your own words. Write the request the way you'd tell a colleague, and attach a screenshot or a BRD if you have one. Your agent takes it from there. | Watch every request move. Each item moves through Understand → Decide → Change → Verify → Remember, and its % is computed from evidence. |
| Get the app running with your agent. The agent works out how the project really runs, including Docker, vhosts and parent folders, and proposes the services. You approve the exact commands once. Its investigation is read-only and never touches secrets. | Then just press ▶ Start project. One page shows the state, Start/Stop, the apps to open, and a row per service. Shared infrastructure is never stopped by mistake. |
Take the full visual tour → It goes from an empty folder to a running app in 13 screens. الجولة كاملة بالعربي ←
flowchart LR
n(["Node.js ≥ 20.16"]) --> i["npm install -g yallaflow"]
i --> p["cd your-project<br/>yallaflow init"]
p --> a["Open Claude Code or Codex<br/>and describe the work"]
p --> c["yallaflow console<br/>(optional, in your browser)"]
1 · Install the CLI (once per machine). Install it globally or in a tools prefix, never as a dependency of your app. Options →
npm install -g yallaflow
yallaflow --version
2 · Connect a project (once per project), or create a new one from the console's New project wizard:
cd your-project
yallaflow init
3 · Work. Open your agent in the project and describe the work, or open the console:
yallaflow console
init connects Claude Code (CLAUDE.md) and Codex (AGENTS.md) with a small pointer and keeps anything already in those files. Then open Claude Code or Codex in the project and describe the work. These are prompts for your agent, not terminal commands:
Fix invoice totals.
Build an anonymous employee survey platform.
You don't need to say "use YallaFlow", explain the architecture, or ask for specs and tests. Every session orients itself automatically.
On an existing project, the first request comes with one review step. The agent first builds an evidence-backed understanding of the codebase, without changing code. It shows you what it found and what it did not inspect, and waits for you to reply Approve. After that, your original request continues, and every later session and teammate starts from that approved understanding. More on first use →
flowchart LR
v["yallaflow --version"] --> u["npm install -g yallaflow@latest"]
u --> s["yallaflow upgrade status<br/>(read-only)"]
s --> r["yallaflow agent refresh<br/>once per project"]
r --> d["yallaflow doctor"]
d --> c(["Restart yallaflow console"])
YallaFlow never updates itself and never checks the network for versions. You decide when to upgrade:
npm install -g yallaflow@latest # or an exact version: [email protected]
yallaflow --version
# in each project
yallaflow upgrade status # read-only: what, if anything, needs attention
yallaflow agent refresh # only if the agent contract is reported outdated
yallaflow doctor
Ctrl-C, then yallaflow console).Updating in detail → · Upgrading to 1.5.0 → · 1.4.0 → · 1.2.0 → · from 1.0.x: 1.1.0 →
以上为 README 的开头部分。其余内容请阅读原文。 查看完整 README ↗