Menu

A durable SSH workflow for AI coding agents: reconnect, inspect, then reply
📰

A durable SSH workflow for AI coding agents: reconnect, inspect, then reply

DEV Community: fastapi·Mooncake·2 days ago
#hdV0qiNf
#devtofastapi#devto#dev#pane#tmux
Reading 0:00
15s threshold

AI disclosure: I used an AI writing assistant to help structure and edit this article. The recovery workflow below is a recommendation, not a benchmark or a description of MoonSSH’s internal implementation.

An agent is running tests on my development host. I close my laptop and open my iPhone. Which process still owns the work? Which pane is it in? And if I already tapped Send once, did the agent receive my instruction?

I built MoonSSH for this workflow. The persistence underneath it comes from the remote host. You can use the commands below with any SSH client.

1. Put the agent behind a server, not just an SSH connection

On your remote development host, change to your checkout and run:

tmux new-session -s checkout -n agent

Enter fullscreen mode

Exit fullscreen mode

Start your coding agent inside that pane. For unattended work, give it a stopping point: “Run the checkout tests and summarize the failures.…

Continue reading — create a free account

Join HashtagPLUS to read full articles, follow hashtags, vote, and join the conversation.