文章

构建 Murmell 的工程手记

我们在一个目录上运行多个编程代理时必须解决的七个工程难题,每一篇都直接来自实现它的代码。所有测量数据均为我们自己的实测结果。

A browser, a control plane and a sandbox. Output flows right to left; the congestion that starts at the browser has to be felt at the sandbox.browsercontrolsandboxoutputoutputslowholdreceive window closesstop writing

构建远程 PTY 终端流式传输与实时多人协同观察架构

保持浏览器与远程运行代理的可靠连接所需的技术细节,以及处理不当时会悄然发生的两类故障。

Three named people on one canvas, reached by a link. The preview is served through the session; a public address for it is refused.one linkaminaeditingmosseditingjonasviewinglive preview*.preview.murmell.comnever

实时共享 AI 编程会话:只读链接协议与状态同步设计

共享链接必须承载的三项核心数据,以及为什么其背后的实时预览极易产生严重安全漏洞。

Above, a task leaves for an isolated environment and returns as a diff. Below, the agent works inside the shared tree and there is nothing to hand back.a task, delegatedits own environmenta diffthe same agent, insideone directory

本地 Codex 与云端 Codex:在统一共享空间中的多代理协作

为什么选择 Codex CLI 还是云端 Codex 取决于谁在监督,而非算力大小。

Three gates in a row. An agent passes the first two and stops at the third.onboardingtrustapi keya fresh container

容器中首次启动 Claude Code 的三个阻碍与解决方案

通过真实 pty 测量的全新容器中阻塞 Claude Code 代理启动的三道关卡。

Two agents reach for the same file. One reservation is granted; the other is refused at the board.the boardclaude-1codex-2src/types/order.tsrefused

写入前预留文件:解决多 AI 代理在同一仓库的代码冲突

为什么同一个目录中的多个代理必须先声明路径占用,以及当发生违规写入时服务器如何处理。

A locked boundary. The database and the key stand on the control side; the sandbox holds only agents and a working tree.controlsandboxdatabasekeynever crosses/workspace

零信任云沙箱架构:运行 AI 代理的机器不持有凭据与敏感信息

为什么执行模型代码的机器与数据库彻底隔离,以及构建这一隔离架构的技术细节。

A sealed machine with one channel through its wall, carrying frames both ways, and a second attempt refused for want of a token.controlsandboxwith a token403without one

AI 代理在云端沙箱运行的实测发现与工程踩坑记录

托管云沙箱平台所允许和拒绝的操作细节,全部来自真实探针集群的实测数据。