claude-code - 💡(How to fix) Fix Team in-process mode: process residue on disband + agents don't consume task_assignment

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…

The in-process backend for team mode has two related bugs:

  1. Process residue: When a team is disbanded (TeamDelete), child agent processes (claude.exe + associated MCP servers like node.exe) are not killed. They remain as orphans consuming memory indefinitely.
  2. Communication dead-loop: In-process agents spawn, report idle_notification, but never consume task_assignment messages from their inbox. They loop "available" without doing any work.

Root Cause

The in-process backend for team mode has two related bugs:

  1. Process residue: When a team is disbanded (TeamDelete), child agent processes (claude.exe + associated MCP servers like node.exe) are not killed. They remain as orphans consuming memory indefinitely.
  2. Communication dead-loop: In-process agents spawn, report idle_notification, but never consume task_assignment messages from their inbox. They loop "available" without doing any work.

Fix Action

Workaround

Use isolation: "worktree" instead of in-process — worktree agents have proper lifecycle management.

Code Example

5 claude.exe processes left (oldest: 5 days)
9 node.exe MCP MySQL server processes left
RAW_BUFFERClick to expand / collapse

@

Summary

The in-process backend for team mode has two related bugs:

  1. Process residue: When a team is disbanded (TeamDelete), child agent processes (claude.exe + associated MCP servers like node.exe) are not killed. They remain as orphans consuming memory indefinitely.
  2. Communication dead-loop: In-process agents spawn, report idle_notification, but never consume task_assignment messages from their inbox. They loop "available" without doing any work.

Reproduction

Environment: Windows 11 Pro (10.0.26200), Claude Code v2.1.150, no WSL.

  1. Create a team with multiple agents using backendType: "in-process"
  2. Assign tasks via inbox task_assignment messages
  3. Observe: agents send idle_notification repeatedly but never execute tasks
  4. Disband the team (TeamDelete)
  5. Observe: claude.exe and node.exe (MCP) processes remain running

Evidence

Process residue

After disbanding 2 teams (test-team 3 agents, factor-mining-phase1 11 agents):

5 claude.exe processes left (oldest: 5 days)
9 node.exe MCP MySQL server processes left

Communication failure

test-team inbox analysis:

  • team-lead asked researcher and git-checker to run git log --oneline -5
  • Both agents only sent idle_notification (idleReason: "available")
  • Both received shutdown_request (read: true) but never reported results
  • Zero task output produced

factor-mining-phase1 (11 agents):

  • 5 Phase 1 agents: only idle_notifications, zero task output
  • 1 agent (phase2-minute-search): succeeded (only one using Explore type)
  • Remaining agents: only idle_notifications + shutdown responses, no task output

Expected behavior

  1. TeamDelete should terminate all child processes in the process tree
  2. In-process agents should poll their inbox and execute task_assignment messages
  3. Agents should report results back to team-lead inbox

Workaround

Use isolation: "worktree" instead of in-process — worktree agents have proper lifecycle management.

Environment

  • OS: Windows 11 Pro (10.0.26200)
  • Claude Code: v2.1.150
  • Shell: PowerShell 5.1
  • Terminal: VS Code + Windows Terminal
  • No WSL, no MSYS2 @

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

FAQ

Expected behavior

  1. TeamDelete should terminate all child processes in the process tree
  2. In-process agents should poll their inbox and execute task_assignment messages
  3. Agents should report results back to team-lead inbox

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING