Artfical AI / Docs
tCode overview Open tAI Install tCode
How it works

Nothing learned is discarded on exit

Exiting plan mode does not throw away what was learned while it was active. The conversation continues with full awareness of everything tCode read and reasoned about while building the plan.

This means asking it to proceed afterward does not restart the investigation from zero; the actual edits happen using the understanding already built up during planning, applied directly rather than re-derived.

In practice this makes plan mode essentially free in terms of wasted work: even if you end up asking for something different from what was originally planned, the investigation that informed the original plan still informs the revised one.

This carries through even a significant redirection. Asking for a noticeably different approach after seeing the plan does not discard the earlier investigation into how the codebase is structured; it is reused as context for whatever comes next, rather than treated as wasted effort tied to the rejected plan.

The practical effect is that trying plan mode has essentially no downside worth worrying about: at worst, you end up asking for something different from what was proposed, and none of the investigation behind it goes to waste either way.