Limit Rescue Join the list

Guide · checked 2 November 2026

How to switch from Claude Code to Codex mid-task without losing your place

Independent project. Not affiliated with Anthropic or OpenAI.

Claude Code is nearly out of usage and you are halfway through a task. You have Codex. Can it just pick up where Claude left off?

Not quite, and that is worth knowing up front. Claude Code and Codex cannot open each other's conversations. Each tool keeps its own transcripts in its own format. What you can move is context: your files, the git state, and a good note about what is done and what is next. Do that well and the switch takes a couple of minutes.

The short answer

Write a "where I left off" note, commit your work in progress, then start Codex in the same folder and tell it to read the note.

Where Limit Rescue fits: it does the note and the launch for you. It warns you at 80% and 95% of your usage limit, writes a "Where I left off" note, and limitrescue resume codex continues the work in Codex. It also goes the other way with limitrescue resume claude. It works with Claude Code and Codex, runs locally, and makes no network calls. It is in early access, not public yet.

Join the list

Limit Rescue v0.1.0 is not on npm yet, has only been tested on macOS so far, and the early-access list is on limitrescue.com. The manual steps below work today with no extra tools.

The manual way

This is the pattern, based on PonyMux's guide (How to hand off a task from Claude Code to Codex), which is a competing tool but has a clear write-up of the general approach.

1. Save your work

Run git status. Commit what works to a work-in-progress branch, or stash it. The next agent will read your files, so you want them in a state you can describe.

2. Get a handoff note while you still can

Ask Claude Code, while it still has usage, to write the note. It knows the dead ends. Something like:

Write a handoff note to HANDOFF.md for another coding agent that will continue this work.
Include: the goal, what is done and how it was verified, what is left in order,
decisions we made and why, the files that matter, and anything uncommitted.
Be specific. Skip anything the next agent can read from the code.

That prompt is adapted from the PonyMux guide. The five things a useful note answers there are: the goal, what is done and how it was checked, what is left in order, the decisions made and why, and which files matter and what is uncommitted.

If Claude Code has already stopped, write the note yourself from git diff. It is the same five questions, just a bit more work. Claude Code blocks requests until the reset time once a limit is hit (see what to do when Claude Code hits its limit), so you cannot ask it after the fact.

Read the note before you hand it over. It is the only thing the next agent knows about the conversation.

3. Start Codex in the same folder

Stay in the same working directory so paths and project instructions line up, then pass the note as the first prompt:

codex "Read HANDOFF.md and continue the work described there."

PonyMux's guide says the Codex CLI takes an initial prompt as an argument and starts an interactive session with it. Watch the first few steps. If Codex starts redoing something the note says is finished, add a line to the note.

4. Share your project rules

Claude Code reads CLAUDE.md and Codex reads AGENTS.md. The same guide suggests keeping one file and pointing the other at it. For example, a CLAUDE.md containing only @AGENTS.md makes Claude Code read your AGENTS.md. Or, if your rules live in CLAUDE.md, the guide says you can add project_doc_fallback_filenames = ["CLAUDE.md"] to ~/.codex/config.toml so Codex falls back to it. Check the current Codex and Claude Code docs for the exact behavior on your versions.

5. Know what does not carry over

From the same guide:

Things to check on the Codex side

Switching only helps if Codex has room.

Quick answers

Can Codex open my Claude Code conversation?

No. Claude Code and Codex cannot open each other's conversations, because each keeps its own transcripts in its own format. What you can move is context: your files, the git state, and a note about what is done and what is next.

What should a handoff note say?

Five things: the goal, what is done and how it was checked, what is left in order, the decisions made and why, and which files matter and what is uncommitted.

Does starting a new Codex chat reset my Codex usage?

No. OpenAI's Help Center says starting a new Codex chat does not reset your usage limits.

How does Limit Rescue help with the switch?

Limit Rescue v0.1.0 warns at 80% and 95% of your usage limit, writes a "Where I left off" note, and limitrescue resume codex continues the work in Codex. limitrescue resume claude goes the other way. It is a local tool that makes no network calls, and it has only been tested on macOS so far.

The Limit Rescue way

Here is the same flow with Limit Rescue.

  1. Work as usual in Claude Code or Codex. Limit Rescue watches your usage on your machine.
  2. At 80% it warns you. At 95% it warns you again.
  3. At 95% it asks the running agent for a "Where I left off" note for the work in progress. If the agent does not answer, it writes a plainer note from the transcript and marks it as a guess.
  4. Run limitrescue resume codex to continue in Codex, or limitrescue resume claude to continue in Claude Code.

The point is that the note exists before you need it, so you are not writing one in a hurry after the wall.

limitrescue resume starts Codex or Claude Code, which talk to their own servers as normal. Limit Rescue itself makes no network calls.

What it does not do yet

Other tools in this space take different approaches, for example resuming the same agent after a reset, or handing off to several other agents. See Claude Code and Codex limit tools compared for an honest side-by-side.

Join the list

Limit Rescue is an independent project and is not affiliated with or endorsed by Anthropic or OpenAI.

Sources