Anthropic's September 2025 Claude Code update added checkpoints for recovering from an unhelpful coding direction. The feature saves code state before Claude's changes and lets the user restore the code, the conversation or both.
That gives developers a recovery point when delegating exploratory work, such as a broad refactor. Its boundary matters: checkpoints cover Claude's edits, excluding user edits and changes made by shell commands. Anthropic recommends using them together with version control rather than as a complete rollback of everything that happened during a session.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
agent hook — A script the harness runs automatically at a fixed point in an agent's loop — before a tool call, after an edit, on session start.
