sannongwangluo

Tatai — Testing skill for Claude Code

Testing community

How to install Tatai

This entry records only its repository, not the path inside it, so there is no exact command to give. Open sannongwangluo/tatai and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Tatai does

Tatai 塔台 — local-first workbench for one person orchestrating multiple AI coding agents (Claude Code / Codex / any MCP agent) on one project: spec-driven design & plans, persistent context across sessions, evidence-verified progress, codebase graphs. 多智能体协作：一个 MCP 接口，所有 agent 共用一份可追溯的事实。

Alternatives in Testing

AI Contribution Skills (Engram) — branch-pr: clean branch and PR workflow 2.5k ★
Spec Driven Develop — Spec-driven development workflow for AI coding agents: architecture-first planning, task decomposition, GitHub 975 ★
Spy — by spylang - Enforces strict coding conventions with comprehensive testing guidelines, multiple code compilati 678 ★

README

塔台 Tatai

**下载 Windows 版** → [Releases](https://github.com/sannongwangluo/tatai/releases/latest) ｜ [上手](docs/getting-started.md) ｜ [开发](docs/development.md) ｜ [更新日志](CHANGELOG.md) ｜ [![CI](https://github.com/sannongwangluo/tatai/actions/workflows/ci.yml/badge.svg)](https://github.com/sannongwangluo/tatai/actions/workflows/ci.yml)

一个人 + 多个 Agent，持续推进同一个项目。

塔台面向用 Agent 开发项目的非程序员：保存意图、设计书、施工安排、进度和验收证据，让换模型、换会话后的 Agent 有依据接着做。数据全部落在你本机，六图、任务账本、Gate、MCP、备份恢复这些核心功能不连网也能用。当前以作者自用为先，作者自行选择工具、模型和档位。

**English**: Tatai ("control tower") is a local-first desktop workbench for one person running multiple AI coding agents (Claude Code, Codex CLI, Gemini CLI, or any MCP-compatible agent) on the same project. Intent, design docs, build plans, tasks, progress and acceptance evidence live in one auditable event ledger that agents read and write through a single MCP interface. It covers multi-agent orchestration, spec-driven development, requirements traceability, acceptance gates and codebase architecture visualization (six graph views). Context survives model and session switches; "done" is only true with verifiable evidence; changing one document re-verifies only what it affects. Local-first, no cloud required. **Tatai does not start your agents for you and does not approve acceptance gates on your behalf.**

**适合谁**：你本机跑 Agent（Kimi Code / Claude Code / Codex 等），项目就在本地磁盘上，希望不同会话、不同模型接的是同一份事实，而不是各聊各的。

**不适合谁**：想找一个“输入需求就自动写完项目”的一键工具。塔台管理协作事实与执行请求，不负责启动或终止 Agent 进程（那是外部协调器的职责），也不代替你验收。

**快速导航**：[能力总览](#能力总览) ｜ [一个贯穿例子](#一个贯穿例子) ｜ [六张图分别看什么](#六张图分别看什么) ｜ [当前能力与边界](#当前能力与边界) ｜ [深入文档](#文档)

0.4.0 更新（2026-10-09）

本轮把人的**默认入口**换成**交付总览**，并补齐接续取材与提交前预检。逐条能力与如实限制见 [CHANGELOG.md](CHANGELOG.md) 的「0.4.0」条目；发布说明见 [docs/releases/v0.4.0.md](docs/releases/v0.4.0.md)。

交付总览（新默认入口）：打开项目先看到的是交付总览——一页核对本期功能、需求依据、技术验证、非作者审查、未解决问题与运行版本；详细检查与证据按需展开。六图保留用于查结构与定位问题，章节方块的颜色不再充当整项目交付结论。只有完整功能范围核查、功能验证、独立审查与运行版本核对都具备当前有效证据，页面才显示「可以开始人工试用」；缺证据、范围未定、旧版本或读取失败会逐条点名。**技术就绪不等于你已经接受——塔台不代你