Using Openclaw Multi-Agent in TUI for Parallel Tasks
Preface
When using Openclaw, you may encounter the following situation.You are currently working on a complex task and have not finished it yet.
Suddenly, a temporary question comes to mind.
For example, you are discussing with Openclaw how to set up the multi-Agent feature in a Telegram group, and you're halfway through, still encountering issues, and haven't finished. (1)
You temporarily see someone discussing whether Openclaw can "see" bash one-click commands on a web page, and whether it is blocked by security settings. You want to give it a try. (2)
If you insert the discussion of (2) in the middle of the discussion of (1), then,
On one hand, the context of (1) will be brought into the processing of (2), introducing unnecessary information interference and token usage.
On the other hand, when (2) is done and you continue discussing (1), the context of (2) will be brought into (1).
If (2) is a particularly simple question that does not require interaction, you can ask Openclaw to start a subagent and report the result back once it's done.
For instance, "What's the weather today?" or "How much CNY is 100 USD?"
This is also the simplest, fastest, and most convenient method.
* Update: the new version of openclaw supports the /btw command.
But, what if (2) is a relatively complex question that may also require interaction?
Then, we need to use a separate agent to handle (2).
TUI
First, let's see how it looks in the command-line TUI.
(* I know many operations have direct commands, but I'm targeting ordinary beginner users, so I'll try to use natural language to communicate with openclaw as much as possible)
In the SSH command line, use the openclaw tui command to open the TUI.
Send
Create a new agent
Supplementary information
1. Agent ID
agent-new-test
2. Purpose/Name
just test
3. Do you need to bind it to a specific channel?
No, I only use it in TUI
4. Model preference?
Use the current one by default
After a while, the new agent will be created.
Use the /agent command in TUI
Then you'll see in the TUI status bar that the agent has changed
At this point, the content you communicate with the new agent in TUI has nothing to do with the original agent.
* BTW: You can open TUI in different SSH sessions and use different agents
You can also give it a try.
========
The End
========
Postscript
The TUI comes with a built-in /agent command to quickly create a new agent.
/agent new agent name
For example
/agent another-new-agent
========
Posts
