Prunes old tool calls and outputs for Claude Code while retaining surviving text verbatim.
Separately judges whether a tool call and its full output are still needed; code keeps, truncates or drops them.
Avoids rewriting paths, commands and errors into a new summary.
Requests are split to fit budgets, rather than always using one call. Pruning can remove useful context; the animation is not a live API benchmark. README and integration source reviewed at a fixed commit; not independently run or benchmarked by this site.
Prune Claude Code tool history while keeping retained evidence verbatim.
Material analysed：
User/assistant text is retained, but tool results may be truncated or removed. The state sent to Jev abridges tool output, so its decision does not necessarily inspect every original byte.
Inspect the supporting source ↗The pinned README describes estimated token fitting and repeated shared state. These are this plugin’s budgets, not a universal current TypeSafe account quota.
Inspect the supporting source ↗Useful where exact tool output matters more than a summary. Before trusting compaction, test later recall of paths, errors and constraints against both compacted and original histories.
Read README’s three decisions and client.ts transport seam. Compare keepResult above threshold with keepCall-only and neither; check that each call/result pair stays structurally consistent in the documented policy.