Comment by wangsj

https://github.com/vince-0202/acgo

Over the past few weeks, I have been building an AI coding tool in Go. The core loop is straightforward: accept a natural-language instruction, let the LLM interpret intent, then execute coding work through tools such as file read/write, code search, and terminal commands.

As of now, I haven't come across any agent coding tools written in Go, but I have always thought that Go is an excellent language and is very suitable for building any CLI tools.

Currently, I have added harness constraints to the agent by exposing hooks and implementing monitoring during the agent's working lifecycle. I think this will enable a clear division of responsibilities between the agent and the harness. The agent is the smallest execution core, while the harness acts as the execution agent for the agent and imposes constraints on its behavior.