[FEATURE REQUEST] Asynchronous Task Tool Execution

Problem Statement

When Claude Code spawns a subagent via the Task tool, it blocks all user input until completion. This forces users to wait idly during long research tasks or open multiple Claude Code instances (losing context).

Potential Solution

I envision being able to run subagents in the background while continuing to interact with Claude. Something like:

• Maybe an async parameter on the existing Task tool
• Or a simple way to mark tasks as non-blocking
• When the background task completes, results appear inline and Claude finishes any analysis it had started

The exact implementation isn't critical - even a basic "don't block on Task" option would help.

Use Case

User: "Research all authentication patterns in this codebase"
Claude: [Spawns background task] "Research started. What else can I help with?"
User: "Fix the typo in README.md"
Claude: [Fixes typo immediately]
[Later...] "Authentication research complete: [results]"

Benefits

• Continue working while long tasks run
• Maintain conversation context (unlike multiple instances)
• Better workflow for mixed quick tasks + deep research