[BUG] UI Freezes When Spawning Multiple Parallel Task Agents
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
Environment:
- Platform: Windows (win32)
- Claude Code CLI
- Model: Claude Sonnet 4.5
Description:
The Claude Code UI becomes completely unresponsive for 10-60+ seconds when Claude spawns multiple Task agents in parallel in a
single response.
Reproduction Steps:
- Ask Claude to perform multiple independent tasks in parallel (e.g., "Fix these 7 linting errors in parallel, one agent per
file")
- Claude responds with a single message containing 7+ Task tool calls
- UI freezes - no input accepted, no visual updates, appears hung
- After 10-60 seconds, UI becomes responsive and shows "Running N Task agents..."
- Agents run fine in background, UI remains responsive during execution
- Brief freeze again during post-processing ("Hashing..." phase)
Expected Behavior:
- UI should remain responsive during agent spawning
- Show progress indicator: "Spawning agent 1/7... 2/7..."
- Tool calls should be processed asynchronously
- User should be able to cancel, type, or interact with UI during spawn phase
Actual Behavior:
- UI completely freezes during the tool call invocation phase
- No visual feedback that anything is happening
- Appears hung/crashed (though it eventually recovers)
- Duration scales with number of parallel agents (7 agents = ~30s freeze)
Workarounds:
- Spawn agents in smaller batches (2-3 at a time)
- Use sequential spawning across multiple messages
- Accept the freeze and wait it out
Additional Context:
This makes the "run tasks in parallel" workflow nearly unusable for large batches. Users expect parallel execution to be
fast/responsive, but the spawning overhead negates the benefits.
Similar Issues:
- Post-processing phase ("Hashing...") also blocks UI when many files are edited
- Suggests synchronous file hashing after agent completion
Suggested Fix:
- Make tool call processing async in UI
- Stream agent spawn events to UI
- Show progress indicator during initialization
- Process file hashing in background with progress updates
What Should Happen?
Expected Behavior:
- UI should remain responsive during agent spawning
- Show progress indicator: "Spawning agent 1/7... 2/7..."
- Tool calls should be processed asynchronously
- User should be able to cancel, type, or interact with UI during spawn phase
Error Messages/Logs
No error messages - UX hung.
Steps to Reproduce
Fix these linting errors (100's) in parallel, one agent per file (10)
Claude Model
Sonnet (default)
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
2.1.37 (Claude Code)
Platform
Anthropic API
Operating System
Windows
Terminal/Shell
PowerShell
Additional Information
_No response_
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗