reddit · SideProject · 11-02 15:50 55 分 ← 返回列表

After two years of coding with agents, I kept losing track of why things were built, so I made a map that links commits back to the messages behind them

I've been building with coding agents for about two years, and the whole time I felt something was slipping away. When you write code yourself, it sinks in bit by bit. You know why that table exists and why you wrote that class. When an agent writes it, that feeling goes away. The information isn't really gone, though. It sits in your messages, the agent's replies and your commits, just in separate places. So I built cogit to tie them together. Every message you send to yo

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：AI agent 写代码后，'为什么这样写'的决策上下文分散在聊天记录和 commit 里，几小时后自己都理不清，代码变成黑箱。

目标用户：同时并行运行 5-10 个 agent、跨 2-3 个项目的重度 AI 编程开发者，以及需要审计 AI 代码来源的小团队技术负责人。

证据：I run 5-10 agents at once across 2-3 projects. Sometimes a few minutes later I can't remember what I did where. / It's free and open source, with no server and no account. / How do you keep track of why your AI-written code ended up the way it did?

变现：核心 CLI/本地版永久免费开源获客；付费点放在团队层：托管的 commit-决策图谱同步、多人共享与检索、按 PR 自动生成'变更理由'报告、合规审计导出（谁/哪个 agent/哪条指令产生了这段代码），以及 JetBrains/Xcode 等多 IDE 与多 agent（Cursor、Codex、Copilot）适配的订阅版。

MVP：两周内做：VS Code 插件 + 本地 SQLite。监听 Claude Code/Cursor 的会话日志文件，按时间戳和 diff 相似度把每条 assistant 消息与 git commit 自动配对，生成'commit → 触发它的对话 → 被否决的方案'可回溯时间线；提供 commit 详情的'为什么'侧栏和一键搜索（按意图/关键词找历史决策），全程本地无账号。第二天即可加'新会话开始时自动注入相关历史决策'的钩子，形成留存。

竞品：SpecStory、git-ai、Cursor Checkpoints、Basic Memory / Mem0 等 memory MCP、Kiro、Tessl 等 spec-driven 工具

摘要：作者用 coding agent 写了两年代码，发现'为什么这么写'的上下文散落在聊天与 commit 中，于是做了开源工具 cogit，把每条发给 agent 的消息变成'thought commit'并与代码 commit 双向关联，还提供 Shape 图概括大改动。作者自述同时跑 5-10 个 agent、跨 2-3 个项目，几分钟后就记不清改了什么；目前仅支持 Claude Code、免费开源、无需服务器和账号。帖中没有任何付费或收入信号，属于真实但尚未被付费验证的开发者痛点。