Agents invoke CodeX, they need to output logs.
Resolved 💬 4 comments Opened Oct 9, 2025 by bradford54 Closed Mar 20, 2026
💡 Likely answer: A maintainer (gpeal, contributor)
responded on this thread — see the highlighted reply below.
What feature would you like to see?
When calling codex through MCP, it usually takes a long time, and during this period, no logs are output. Also, no return is visible in the calling segment. Sometimes, if it takes more than a few minutes, it becomes impossible to wait because we don't know whether it's still running or dead. We hope that during the running process, the logs of that part can be output, so that the caller can also know it, or it can be output to a file as well. Then we can know whether the service is running or at which part it has reached.
Additional information
_No response_
4 Comments
Are you referring to the Codex CLI calling a tool on a MCP server? If so, which logs are you referring to, specifically? MCP progress notifications?
Hello @gpeal , I'm not sure if this request has been resolved or answered.
Based on this example: https://developers.openai.com/codex/guides/agents-sdk/
The issue is that when the Codex tool mcp-server is run, no information is received about the commands/tools that Codex executes. Depending on the task sent, 10, 15, or 30 minutes can pass without knowing what it's doing.
When working from the IDE plugin (VSCode), the different traces of the Codex tool executions are visible.
Regards
I have the same problem as @hayhu
We are likely going to deprecate the ability to call Codex as an MCP server. In any case, this isn't an area where we are investing, so I'm going to close this feature request.