[Feature Request] Claude Code Sub Agents

[ndduy]

Anthropic released their version of specialized sub-agents.

Do you consider adding it to the project anytime soon?

https://docs.anthropic.com/en/docs/claude-code/sub-agents

[peterkrueck]

Hey,

I've already looked into it and am not yet sure how to integrate them into my workflow. Personally, I haven’t used them yet. What I really like about the standard sub-tasks (not the custom agents) is that the main Claude Code assigns dynamically tasks and personalities to them.

Will think about it. If you've an idea that integrates well and does meaningfully better than the current approach, I'm open to contributions!

[tukusejssirs]

When using subagents, you don’t clutter the main Claude instance context with unnecessary stuff (only the final output of an agent use processed by it).

What I really like about the standard sub-tasks (not the custom agents) is that the main Claude Code assigns dynamically tasks and personalities to them.

With subagents, you can specify when to use an agent in the agent description (then Claude will use them automatically when the conditions match). You can also ask Claude explicitly to use an agent for a particular task.

[peterkrueck]

When using subagents, you don’t clutter the main Claude instance context with unnecessary stuff (only the final output of an agent use processed by it).

What I really like about the standard sub-tasks (not the custom agents) is that the main Claude Code assigns dynamically tasks and personalities to them.

With subagents, you can specify when to use an agent in the agent description (then Claude will use them automatically when the conditions match). You can also ask Claude explicitly to use an agent for a particular task.

Yeah I'm aware of sub agents don't clutter context window, but it's the same for sub tasks too. As said before, they might be really helpful maybe for having purposely built sub agents for frontend or database or deployment. Currently haven't experimented with them yet though.