[Feature Request] Expose raw thinking tokens from extended thinking models
Resolved 💬 3 comments Opened Mar 12, 2026 by SupposedlySam Closed Mar 16, 2026
Feature Description
Please add a feature where I can get raw thinking tokens. 1. It makes it feel faster and I get to learn as Claude explores the codebase, 2. I can know when it's going off the rails or when it's trying to find something in the wrong place before it makes code changes. What I'm looking for is like what I get OOTB from using these same models inside of Cursor's chat pane.
Environment Info
- Platform: darwin
- Terminal: Apple_Terminal
- Version: 2.1.74
- Feedback ID: c29be901-8519-4449-8c5a-0b02991cbb04
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗