Tracking the rest, with what is known about each. Open a dedicated issue when one is picked up.
- OpenCode — claude-replay and AgentGit (native resume) support it; storage format to determine.
- GitHub Copilot CLI — Entire hooks it; log location to determine.
- Pi — Entire and vibe-replay support it.
- Kimi Code —
~/.kimi-code/sessions/<project>/<session>/agents/<name>/ (claude-replay).
- Hermes Agent —
~/.hermes/state.db, SQLite; claude-replay reads it live. Needs Node 22.5+ node:sqlite.
- Factory AI Droid — Entire hooks it.
- ATIF / OpenHands trajectories — not a runtime but a format: Harbor's Agent Trajectory Interchange Format and the OpenHands trajectory datasets on Hugging Face (
nebius/SWE-rebench-openhands-trajectories, nvidia/SWE-Hero-openhands-trajectories). An import adapter would let agit verify, replay and diff benchmark runs. See the interop issue for the export side.
Count today: Entire 8, claude-replay 7, vibe-replay 4, agit 3.
Tracking the rest, with what is known about each. Open a dedicated issue when one is picked up.
~/.kimi-code/sessions/<project>/<session>/agents/<name>/(claude-replay).~/.hermes/state.db, SQLite; claude-replay reads it live. Needs Node 22.5+node:sqlite.nebius/SWE-rebench-openhands-trajectories,nvidia/SWE-Hero-openhands-trajectories). An import adapter would let agit verify, replay and diff benchmark runs. See the interop issue for the export side.Count today: Entire 8, claude-replay 7, vibe-replay 4, agit 3.