›_ Codex Issues
Codex Issues / #17448

Forever waiting for xcodebuildmcp

Open 💬 4 comments Opened Apr 11, 2026 by vipuldelwadia
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What version of Codex CLI is running?

0.120.0

What subscription do you have?

Go

Which model were you using?

gpt-5.4

What platform is your computer?

Darwin 25.4.0 arm64 arm

What terminal emulator and version are you using (if applicable)?

Ghostty

What issue are you seeing?

Looks like it it never finishes waiting for xcodebuildmcp

What steps can reproduce the bug?

Uploaded thread: 019d7bce-17f6-7302-b790-01849b3ba25e

What is the expected behavior?

_No response_

Additional information

_No response_

View original on GitHub ↗

4 Comments

github-actions[bot] contributor · 4 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #17444
  • #17143
  • #17024

Powered by Codex Action

etraut-openai contributor · 4 months ago

If you're able to repro this, please use /feedback to upload logs and session details, then post the thread ID here.

jounih · 3 months ago

Adding a fresh data point from macOS / Codex CLI 0.128.0.

Primary symptom: Codex CLI hangs during MCP scanning and does not proceed until the MCP scan is manually cancelled. After cancellation, Codex continues, but xcodebuildmcp helper processes can remain running.

Environment:

  • macOS 26.5, Darwin 25.5.0, arm64
  • Codex CLI 0.128.0
  • Active Codex process was codex resume ...

Relevant setup:

  • ~/.codex/config.toml initially had an explicit [mcp_servers.XcodeBuildMCP] entry:
  • npx -y xcodebuildmcp@latest mcp
  • The OpenAI-curated build-ios-apps@openai-curated plugin was also enabled.
  • That plugin declares another MCP server named xcodebuildmcp:
  • npx -y xcodebuildmcp@latest mcp
  • with XCODEBUILDMCP_ENABLED_WORKFLOWS=simulator,ui-automation,debugging,logging

Observed:

  • Codex effectively had both XcodeBuildMCP and xcodebuildmcp available from two sources.
  • The CLI got stuck during MCP scan until the scan was cancelled.
  • ps showed many stale process pairs left behind:
  • npm exec xcodebuildmcp@latest mcp
  • node ~/.npm/_npx/.../node_modules/.bin/xcodebuildmcp mcp
  • Removing one source fixed codex mcp list returning quickly.
  • I switched to curated-plugin-only because keeping the build-ios-apps skill is useful, but even the bounded verification command spawned two npm exec xcodebuildmcp@latest mcp parents and two node ... xcodebuildmcp mcp children, which had to be killed manually afterward.

This may overlap with #17574, but from the CLI user perspective the main failure is: hangs at MCP scan until the user cancels it.

Expected behavior:

  • Codex should not hang indefinitely during MCP scan/startup on xcodebuildmcp.
  • Codex should either deduplicate/warn when the same MCP command is configured both explicitly and through a curated plugin, or avoid starting both.
  • Helper subprocesses spawned during MCP discovery/listing should be torn down when discovery is cancelled or completes.
jounih · 3 months ago

Additional user-impact detail: this is not just a slow scan. Codex often starts MCP scanning on its own during normal use, so the CLI gets stuck frequently without an explicit MCP action from the user.

In practice I have to manually cancel the MCP scan, even after it has been sitting there for a very long time, potentially around an hour, and then tell Codex not to use MCPs for the session. Otherwise the CLI remains blocked at MCP scanning