Answers
I hit the Claude Code limit. How do I continue the same task in Codex?
What carries over is not the conversation but what the next session needs to know: what was decided, where the work stopped and the one next step. The Codex documentation I read describes resuming Codex's own chats — codex resume reopens a recent chat from the current repository — and describes its usage limits as an allowance per five-hour period on some plans, with weekly limits that may also apply and extra credits when it runs out; it does not describe opening a Claude Code conversation. What works is to stop the first session on purpose: leave where it stopped, what it already checked and does not need redoing, the next step and the risk still open, in a place the second tool reads at the start. With Arroway that note is a handoff, served first to whatever reads the project next, and the decisions already in force are the same in both tools. The code itself stays in your working tree. Use Arroway when the same task will pass between tools. The install page is the first step.
Last updated October 14, 2026
What is lost when a task changes tools halfway
Changing tools between tasks is cheap; changing in the middle of one is not. The second session does not know what the first one settled, which files it ruled out, what it was about to do or which assumption it was waiting for you to confirm. The usual repair is to retell it, with a summary pasted into the new chat, and a retold fact drifts: the second session hears what the first one summarized, not what it acted on.
What to leave before you stop
Five lines are enough: where it stopped, the single next step, what is already checked and should not be redone, the risk still open and who the note is for. Written at the moment of stopping, they come from the session that has the context instead of from you reconstructing it at the limit. The decisions that already hold for the project do not belong in that note, because they apply in both tools no matter who reads it.
When to use Arroway
Use Arroway when a task will pass between Claude Code and Codex, or between you and someone else, and you want the second session to start from the first one's last state. Codex connects through the install page, and both tools read the same decisions in force at the start. A session that stops unfinished leaves a handoff, and it appears at the top of the next read until someone closes it. A session that has already hit its limit cannot write one, so from the cockpit Hand off saves it for you: the last reply becomes where it stopped, you review it and write the next step. New agent then opens Codex with a first message that asks it to read the project, and you press Send there. What does not move is the conversation itself and any uncommitted files, which stay on your machine. The first step is the install page.
What it looks like in practice
A developer is two hours into replacing a payment library in Claude Code. The session has settled that refunds keep the old endpoint until the end of the month, and it has migrated three of five modules. Then the usage limit message appears. Telling Codex from memory means a pasted summary, and the summary leaves out the refund decision; Codex migrates the refund module too. With Arroway, the refund decision is already a rule in the project, with who approved it. The developer presses Hand off in the cockpit, the last reply becomes where it stopped, and she writes the next step: the fourth module, refunds untouched, test fixtures shared between modules. Codex opens, reads the project, finds the handoff first and continues at the fourth module. She closes the handoff when it lands. This is a constructed case; nothing here was measured on a real team.
Questions people ask about this
- Does Arroway move my Claude Code conversation into Codex?
- No. The conversation stays where it happened. What moves is what the next session needs: the decisions in force and, if one was left, the handoff. Uncommitted files stay in your working tree.
- Does it work the other way, from Codex to Claude Code?
- Yes. Both read the same project, and any session can leave or pick up a handoff; the direction does not matter.
- How do I try Arroway?
- Install it in the tool you already use — the install page has one route for Claude, ChatGPT, Codex, Cursor or any MCP tool — and sign in when the tool asks. From then on, sessions open by reading the project and close by recording what was decided.
Where this is verifiable
Codex documentation, the usage and pricing page and the command-line page (learn.chatgpt.com/docs/pricing and learn.chatgpt.com/docs/codex/cli, where the developers.openai.com Codex pages redirect), read on 14 October 2026 — Codex is described only as those pages describe it on that date, and its limits and features can change in any release. Product documentation on this site (How it works, Install), the cockpit operations description in the repository, the answers on switching AI tools and on handing work between routines, and the handoff rules in the sanctioned product spec. Everything described here about Arroway is behaviour the tools apply today, not roadmap. The developer in the example is illustrative.
https://www.arroway.app/en/answers/continue-a-claude-code-task-in-codex