Keep the whole agent herd in view
Run Herdr beside your coding agents today and prepare the host for TalkTerm’s planned optional native overview.
Herdr is an optional terminal multiplexer built around coding agents. It keeps panes alive and shows which agents are working, blocked, done, or idle.
One host. One herd. Normal SSH still works without it.
What Herdr adds
Herdr runs where the repositories, credentials, and agents already live. It gives every workspace real terminal panes and a shared view of agent state. Closing the client detaches you; it does not stop the background server or its panes.
You do not need Herdr for TalkTerm’s normal SSH terminals. Treat it as an optional coordination layer when several agents make raw terminal tabs hard to scan.
Put it where the work lives
Install Herdr on the remote host, then start it from the project directory. Use the same restricted operating-system user that owns the repositories and agent credentials. Keep the Herdr socket local to that machine; remote access should still cross your verified SSH connection.
herdr --version
herdr
herdr session list --json
herdr api snapshotKeep the boundary boring
- Expose SSH, not Herdr’s local socket.
- Use one restricted user for agent work.
- Confirm every new SSH host key and fail closed when it changes.
- Stop a Herdr session only when you really want its pane processes to exit.
What TalkTerm will use
The planned integration uses Herdr’s machine-readable session snapshot through the already verified SSH host. That keeps agent state, actions, and terminal access on one private connection instead of publishing another dashboard or service.
Until that native view ships, the same host remains useful as a normal TalkTerm SSH destination and Herdr remains fully usable from a terminal.