[BUG]

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?

On one specific Windows 11 (AMD) machine, the Claude Code agent fails on every turn: it connects to the An
thropic API and exchanges data, but produces zero output (empty stdout AND stderr) and exits after ~3
seconds.

In the Claude Desktop "Code" tab this appears as a spinner followed by:
[CCD CycleHealth] unhealthy cycle for local_<id> (1s, hadFirstResponse=false, reason=no_response)

The same failure reproduces with the bundled standalone binary: claude.exe -p "hi" exits in ~3s with emp
ty stdout and stderr, even with --output-format stream-json --verbose. claude --version works fine.
Cowork (VM mode), Chat, and the Linux/WSL Claude Code CLI all work on the same machine and account; the sa
me account also works on a different PC.

I packet-captured the failing request (Windows pktmon) and confirmed the network path is fully healthy,
so this looks like an internal failure of the Windows agent binary in this environment.

What Should Happen?

claude -p "hi" should print the assistant's response (or at minimum emit a structured error on stderr /
in stream-json), and the Desktop "Code" tab should return a normal response instead of reason=no_response
.

Error Messages/Logs

--debug log consistently stops here, then ~3s of silence, then exit (buffers appear unflushed):

[DEBUG] Git remote URL: null
[DEBUG] No git remote URL found


Desktop main.log:

Session <id> query iterator completed
[CCD CycleHealth] unhealthy cycle for local_<id> (1s, hadFirstResponse=false, reason=no_response)


pktmon (TCP/443, all components) for claude.exe -> api.anthropic.com (160.79.104.10:443):
- 2 connections established (SYN / SYN-ACK)
- ~90 PSH data packets exchanged bidirectionally
- server-initiated FIN arriving at the physical NIC (genuine remote close)
- ZERO packet drops at any component (no firewall/EDR/VPN interference)

Steps to Reproduce

1. On the affected Windows machine, run the bundled agent binary directly (desktop not involved):

"%APPDATA%\Claude\claude-code\2.1.177\claude.exe" -p "hi"

1. Observe: the process runs ~3s, then exits with empty stdout AND empty stderr.
2. Same result with --output-format json and --output-format stream-json --verbose (still zero stdout

— abnormal, since stream-json should emit an init event immediately).

1. In Claude Desktop, the "Code" (local agent) tab shows a spinner and reason=no_response for every mess

age.

Claude Model

None

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.177

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

Ruled out on this machine:

• Network/security: pktmon proves the API conversation actually happens and no filter drops/injects packet

s. GlobalProtect / Windows Defender / Windows Firewall all clean; CrowdStrike Falcon fully uninstalled; a
Defender path+process exclusion made no difference.

• Auth: OAuth token valid (not expired; Max plan; scopes include user:inference, user:sessions:claude_code

).

• Config: reproduced with CLAUDE_CONFIG_DIR pointing at an empty temp dir containing only .credentials.jso

n (no CLAUDE.md / settings / MCP) — identical failure.

• Binary: claude --version prints 2.1.177 (Claude Code) correctly.

Environment: Windows 11 Pro, AMD CPU. Claude Desktop MSIX build 1.13576.0 (also reproduced on the older Sq
uirrel build). WSL2 Ubuntu installed; working dirs often on \\wsl.localhost\... Detected shell: C:\Program
Files\Git\bin\bash.exe. The repro above uses a normal Windows working directory, so it is not specific to
the WSL UNC path. The ~3s runtime is suspiciously consistent (possible internal first-response timeout).

Happy to provide the full pktmon capture, ~/.claude/debug logs, and the desktop main.log on request.

anthropic_bug_report_claude_code_win_no_response.md