Where does the bug appear (feature/product)?
Cursor CLI
Describe the Bug
I am attempting to use the Cursor CLI tool, in print mode, with the “auto” model, but keep receiving a confusing error message.
Here is an example for the command I am using:
cursor agent --print --force --trust --workspace /path/to/workspace --model auto "$(cat "$PROMPT_FILE")" 2>&1 | tee /tmp/cursor-agent-capture.txt
Here is the confusing error message:
Cannot use this model: "auto". Available models: auto, composer-1.5, composer-1, gpt-5.3-codex, gpt-5.3-codex-low, gpt-5.3-codex-high, gpt-5.3-codex-xhigh, gpt-5.3-codex-fast, gpt-5.3-codex-low-fast, gpt-5.3-codex-high-fast, gpt-5.3-codex-xhigh-fast, gpt-5.3-codex-spark-preview, gpt-5.2, gpt-5.2-codex, gpt-5.2-codex-high, gpt-5.2-codex-low, gpt-5.2-codex-xhigh, gpt-5.2-codex-fast, gpt-5.2-codex-high-fast, gpt-5.2-codex-low-fast, gpt-5.2-codex-xhigh-fast, gpt-5.1-codex-max, gpt-5.1-codex-max-high, opus-4.6-thinking, gpt-5.2-high, opus-4.6, opus-4.5, opus-4.5-thinking, sonnet-4.6, sonnet-4.6-thinking, sonnet-4.5, sonnet-4.5-thinking, gpt-5.1-high, gemini-3.1-pro, gemini-3-pro, gemini-3-flash, gpt-5.1-codex-mini, grok, kimi-k2.5
I am using “auto”, and “auto” is literally the first available model in the list of valid models, yet I still receive the error.
The same prompt, when used in interactive mode, works just fine. Only when used in print mode does it fail immediately with the confusing error.
This seems to be a bug.
Steps to Reproduce
-
Have some non-auto usage available (at least a few percent) and a high amount of auto usage available (at least 50%).
-
Call the CLI tool with the
--printoption and the--model autooption, redirecting the output to a log file. -
Notice that you receive an error.
Expected Behavior
I expect that since “auto” is the first available model, that I should be able to use it without incident.
Operating System
MacOS
Version Information
CLI Version 2026.02.13-41ac335
For AI issues: which model did you use?
I used “auto”
For AI issues: add Request ID with privacy disabled
I cannot get the Request ID using the CLI agent outside of interactive mode. There should be bursts of 7 failed request (they fail within 2 seconds) in a row today around 11:45am EST for my account (if that even helps).
Additional Information
It seems to work just fine in interactive mode, but when in print-mode (I also use --trust) it will not allow using auto.
Does this stop you from using Cursor
Yes - Cursor is unusable