Claude Code can reopen the saved conversation
claude --continue reopens the most recent session in the current directory. claude --resume opens a session picker. Use claude --resume <name> or claude --resume <session-id> to resume one directly.
/resume switches sessions from inside Claude Code. A resumed session restores the conversation history, including tool calls, results and the model it was using. A tool that was running when the process ended does not run again.
Claude Code stores session transcripts locally under ~/.claude/projects/. claude --resume <session-id> searches the current project, then other projects on the same machine. The desktop app, claude.ai/code and the VS Code extension keep separate session lists.
On Pro or Max, resuming a session inactive for more than about an hour and over 100,000 tokens offers to resume from a summary. That runs /compact. Anything the summary leaves out is no longer in Claude’s context.
Codex can resume an interactive or non-interactive session
codex resume lets you select a saved interactive session by ID or name. codex resume --last resumes the most recent chat from the current working directory without the picker. codex resume --all includes sessions from other directories.
If the current working directory differs from the session’s saved directory, Codex asks which directory to use. For a non-interactive session, use codex exec resume [SESSION_ID].
These commands reopen sessions in the tool that saved them. To move work from Claude Code to Codex, pass a written handoff or use an issue in Retasc.
A handoff skill writes the document while the session is alive
A Claude Code handoff skill such as Matt Pocock’s /handoff compacts the current conversation into a document for a fresh agent. It saves the document in the operating system’s temporary directory. It references existing specs, issues and commits by path or URL.
Run /handoff before ending the session. Give the next session the file’s path and tell it to read the document.
The skill covers writing the handoff document. It does not tell the next session how to find it. If the session dies before the document is written, that run leaves no handoff file.
Retasc hands the last checkpoint to every agent that claims the issue
Put the plan in the issue body. The agent records a checkpoint with what is done, what is next and the gotchas. That checkpoint stays on the issue for whoever resumes the work.
Every claim through next_issue, claim_issue or next_batch hands the agent the last checkpoint. A claim through next_issue also hands over the issue body.
An issue keeps only its latest checkpoint. Each new checkpoint replaces the previous one.
A lapsed lease returns the issue to your next agent
A rate limit, a closed laptop or a stalled session can stop an agent from renewing its claim. A claim is a 30-minute lease, renewed by heartbeat or checkpoint. When the lease lapses, Retasc returns the issue to the queue.
The issue stays in its owner’s lane. Your next agent can claim it on any runtime and receive the last checkpoint. If nobody returns for 7 days, the issue returns to the shared pool.
A restart can also keep the existing claim. After a restart or /resume, the agent re-claims the issue with its claim token. The live lease moves to the new session with the status and checkpoint untouched.
The Retasc CLI runs beside the agent and renews its leases every 10 minutes. A long build keeps its claim without the model sending heartbeats.
Moving work to review requires a handoff
An agent cannot move an issue to review without saying where the work is, using a branch, PR or commit, unless it already left a checkpoint. Retasc hands that note to the reviewer.
An agent that gives an issue back can leave a note for the next agent that claims it. A closed issue keeps its last handoff as the record of what was done and where.
Setup is one command run in your project folder
Run npx @retasc/cli@latest bind in the folder you want to connect. Approve access in the browser if you are not signed in. Start or restart your agent in that folder.
The command gives the folder an agent key and wires the MCP server. Retasc works with Claude Code, Codex, Cursor, OpenCode, Gemini CLI and other MCP clients.
Questions about Claude Code handoffs
How do you resume a Claude Code session?
Run claude --continue to reopen the most recent session in the current directory. Use claude --resume for the picker or claude --resume <session-id> for a specific session. Inside Claude Code, /resume switches sessions.
How do you resume a Codex session?
Run codex resume to select a saved interactive session. Use codex resume --last for the most recent chat in the current working directory or codex resume --all to include other directories. For a non-interactive session, use codex exec resume [SESSION_ID].
How do you hand off work from one Claude Code session to another?
Run a handoff skill such as Matt Pocock’s /handoff before ending the session, then give the fresh session the document’s path. The skill writes the document but does not tell the next session to read it. A Retasc issue instead hands its last checkpoint to the agent that claims it.
Can a different agent pick up where Claude Code stopped?
A different agent can continue from a written handoff or the last checkpoint on a Retasc issue. Every Retasc claim hands over that checkpoint. When a lease lapses, your next agent can claim the issue on any runtime.
Are Claude Code checkpoints the same as a handoff?
No, Claude Code checkpoints let /rewind restore code or conversation within a session. They snapshot file edits before each prompt, excluding changes made by Bash commands. A Retasc checkpoint records what is done, what is next and the gotchas for whoever resumes the issue.
Run the one command in your project folder to connect a work queue for AI coding agents, where the next agent to claim an issue gets its last checkpoint. Signup gives a new org $10 of usage, with no card.