# CodexLoom

A working environment for long-running Codex agents that gives each agent a durable domain responsibility, organizes agents into a governed team, and exposes mature capabilities to customers and collaborators through interface agents.

CodexLoom is built on top of Codex and does not reimplement its agent runtime or duplicate thread history. Instead, it turns a Codex thread into the ongoing workspace of a long-running domain agent, then adds durable identity, a profile, team relationships, bounded coordination, human governance, and governed external delivery around it.

An agent in CodexLoom has a stable ID, name, profile, and primary thread, resumable through Codex Desktop, Mobile, or the CodexLoom web interface. Other agents collaborate through bounded messages and topics and consume explicit managed artifact handoffs, rather than resuming another agent's primary thread directly. External collaborators can work through governed interface agents connected to platforms such as Feishu, Slack, and Parall. As repeated work reveals stable boundaries, an owner can divide a single agent's responsibility among several agents and declare how they collaborate, assigning roles such as lead, internal agent, or interface agent.

It is written in Go, with its canonical product documentation maintained in Simplified Chinese and translated into English.

Website: https://github.com/yan5xu/codexloom