Switch tools. Keep everything.
Every tool keeps its own history to itself, so moving tools means losing yours.
Note 1 — nothing is migrated
It just reads what each tool already writes. Nothing gets converted, and nothing gets moved.
Fig. 1 — already on diskSix surfaces, five tools
~/.claude Claude Code sessions, plans, tasks, history opencode.db OpenCode sessions and todos ~/.codex Codex sessions ~/.gemini/antigravity-cli Antigravity sessions ~/.cursor Cursor CLI sessions ~/.config/Cursor Cursor IDE sessions
What you lose when you move
Each assistant only sees its own history, never any other tool's.
01 · The loss
The reasoning is what actually disappears
The architecture debate you had, the approach you tried and rejected, the one command that finally worked.
02 · The trial
A new tool starts already knowing your history
It reads the same memory from its very first run, so trying it out tests the tool, not your patience.
03 · The answer
It covers the tools you stopped using, too
The answer spans all five, including the one you abandoned. History stops being a reason to stay somewhere.
Using a tool we don't read yet?
Tell us which one. The next one gets built because someone actually asked.
Or write it yourself: a tool is one adapter file plus one line in an index. docs/ADDING_A_TOOL.md.