Windows: /mcp prints taskkill “SUCCESS” process termination logs
Resolved 💬 12 comments Opened May 3, 2026 by RE-codes Closed May 10, 2026
What version of Codex CLI is running?
codex-cli 0.128.0
---
What OS and shell are you using?
- Windows 11
- PowerShell
---
What steps can reproduce the bug?
- Configure an MCP server (e.g. Context7)
- Start Codex CLI
- Run
/mcp
---
What is the expected behavior?
/mcp should display MCP server status without OS-level process logs.
---
What is the actual behavior?
SUCCESS: The process with PID 14276 (child process of PID 5280) has been terminated.
SUCCESS: The process with PID 5280 (child process of PID 14200) has been terminated.
SUCCESS: The process with PID 14200 (child process of PID 14748) has been terminated.
SUCCESS: The process with PID 14748 (child process of PID 2024) has been terminated.
appear every time /mcp is run.
---
Additional context
- Context7 MCP server works correctly; this is cosmetic noise
- Reproduces with:
npxdirectlycmd /c npx- custom
env startup_timeout_ms- This appears to be Windows-specific (not observed under WSL2)
---
Impact
Low functional impact, but degrades CLI UX—especially when /mcp is used frequently.
12 Comments
I encountered the same problem.
Same issue:
<img width="1068" height="468" alt="Image" src="https://github.com/user-attachments/assets/897f756a-ae5f-4513-9bbe-84e2c297a122" />
<img width="752" height="555" alt="Image" src="https://github.com/user-attachments/assets/600fa9bc-6a7c-4059-9787-ed2f50eafe02" />
Relevant code:
The Windows termination path currently appears to invoke taskkill without redirecting its stdio, see
codex-rs/rmcp-client/src/stdio_server_launcher.rs:Since no stdio is specified, taskkill inherits Codex's stdout/stderr. As a result, taskkill status messages are printed directly into the terminal used by the TUI.
Possible fix:
Redirect the taskkill process stdio to null:
Optionally, also set
CREATE_NO_WINDOWon Windows:I prepared a possible fix:
EurekaMXZ/codex@134409dHowever, I am unable to create a Pull Request for this commit, as it only allows invited collaborators to do so.
If this approach matches the intended direction, I’d be happy to open an invited PR.
@etraut-openai Please fix this bug. Thank you so much!
<img width="895" height="318" alt="Image" src="https://github.com/user-attachments/assets/9837e2e1-ad71-4a62-8c34-513bde135a37" />
Same problem
Same. thread ID 019df5b9-d5f8-77a3-8629-36c48d1879b1
Same... version
codex-cli 0.128.0<img width="1214" height="182" alt="Image" src="https://github.com/user-attachments/assets/692a24ee-18ce-438a-a2a6-8ad6f67dc205" />
@rhan-oai Please fix this bug. Thank you so much!
@fcoury-oai Please fix this bug. Thank you so much!
Fixing this bug could still take a long time; this repository does not accept PRs from external contributors, and there are currently over 3,000 open issues.
Currently, in the latest version of Codex, if this issue occurs, you can resize the terminal window; this triggers a re-layout of the Codex TUI, thereby clearing the stdout output by
taskkillthat has leaked into the TUI.As of today (May 10) the issue is still present for me and triggers regularly
<img width="1052" height="599" alt="Image" src="https://github.com/user-attachments/assets/4b8eca3f-cf78-4f48-8eff-b070de78466e" />