Skip to main content
Forking hands a chat that started on one harness (for example Claude Code) to another harness (for example Codex or Cursor) in the same workspace. The original chat is left exactly as it was. Forks are instant and deterministic: no model summarizes the conversation.

​
How it works

  1. Replicas reads the source chat’s transcript, strips tool noise, redacts secrets, and saves the sanitized transcript inside the workspace. Chats larger than 32 MB are sanitized from their most recent part only; the complete raw event log stays on the workspace for the agent to grep, and the handoff says so.
  2. A new chat opens on the target harness. Its first message says where the chat came from, links the sanitized transcript and the raw event log so the agent can read or grep them, and inlines the most recent turns.
  3. If the source chat’s last request was never answered, it is resent verbatim so the new harness finishes it.
  4. The forked chat shows which chat and harness it came from.
Workspace files and git state are shared, so the new harness continues from the same code.

​
Fork a chat manually

Click + next to the chat tabs and choose Fork this chat to; only harnesses with credentials in the workspace are listed. Pick the model and effort (your defaults for that harness are preselected), and the forked chat opens as a new tab. Sleeping workspaces wake first.

​
Automatic fallback

When a harness hits its usage limit and no credential fallback can continue the turn, Replicas can fork the chat for you.
  1. Open Settings → Preferences → Automatic Harness Fallback.
  2. Turn on Fork chats automatically.
  3. Add harnesses in the order to try. Only harnesses with credentials connected are used, and a chat never falls back to a harness already in its fork chain.
Automatic forks use the model and effort from your agent defaults for the target harness, and resend the interrupted request so the new harness picks up where the previous one stopped.