One ticket on Claude, the next on Codex.
Both plans used in step.
You pay for Claude and for ChatGPT, yet one of them does all the work until it hits its usage limit. AgentsRoom shares your new agents between the two, in turn: one ticket out of two on Codex, or 10% of the Claude session then 10% of the Codex week.
It is a quota rule, Alternate providers, set up in a few clicks in the Usage panel. The same rules can also react to a threshold: switch to Codex when Claude is almost full, pick a lighter model, block scheduled launches, or move running agents to an account with room left.
Alternate providers, Number of launches mode: Claude ×1 then Codex ×1. Each new agent goes to the provider whose turn it is, and an agent already running keeps its tool.
Many developers now pay for two coding subscriptions, often Claude for Claude Code and ChatGPT for Codex. In practice they open one CLI out of habit, run it until the usage limit stops them, then fall back on the other one. The first plan runs out on a Wednesday, the second barely moves, and its week resets having paid for nothing.
Using Claude Code and Codex together, on purpose, means deciding up front which agent goes where. AgentsRoom already reads every quota bar of every CLI you are signed in to, in the Usage panel of the footer. Since 6 October 2026 a quota rule can use those readings to deal out new agents: Claude, then Codex, then Claude again, in the proportion you set.
Both CLIs keep running side by side in the same window, each agent in its own terminal. You keep the strengths of each tool and spend both quotas instead of one.
Two ways to alternate
Pick how a turn is measured. Either way, only brand-new agents are shared out.
Number of launches
Each row of the rotation takes its launches in a row, then hands over. Claude ×1 then Codex ×1 is one ticket out of two on each. Claude ×2 then Codex ×1 gives two for one. Up to twenty launches per turn, and up to six rows.
Claude ×1, Codex ×1: tickets 1, 3 and 5 on Claude, tickets 2, 4 and 6 on Codex.
Quota used
Each row keeps the launches until its own gauge, session or week, has moved by its share, then hands over. A gauge that resets also ends the turn. A plan that publishes no session window, like some Codex plans with only a weekly limit, is measured on its week.
10% of the Claude session, then 10% of the Codex week, then back to Claude.
Or two Claude accounts
The rotation also works inside one provider. Put Claude twice and pick a different account on each row: the field appears as soon as a provider has two signed-in accounts. In Quota used mode each row reads the gauge of its own account, for example 20% of the Personal session, then 20% of the Work session.
Which launches, which project
Limit the rule to Backlog tickets, Scheduled agents, Webhook calls or Manual launches, or keep Every launch. Pick one project or All projects. Each row can force a model, or keep Agent's model: the one the agent would have started with, from its settings, the ticket or the project.
What alternation never does
- It never moves an agent that is already running. A restarted or resumed agent comes back on the provider the rotation gave it, with its conversation, and does not count as a new turn.
- It never writes on the agent's saved settings: the choice applies to that launch only.
- It never sends a launch to a full provider. A row at 99% or more skips its turn, and when every row is full the launch goes through unchanged.
- It never notifies you at each launch. The agent's terminal prints one line naming the rule and the provider it runs on, and the rule card reads Next launch: Codex.
What else a quota rule can do
One condition on a quota bar, one action. Rules apply in the order of the list, and the first satisfied one wins.
Alternate providers
No threshold: the new launches of one provider are shared in turn between two to six rows, providers or accounts.
Change provider, model or effort
While the condition holds, new launches start on another provider, model or reasoning effort. Fable can go to GPT-6 Astra and Sonnet to GPT-6 Luna, one row per source model.
Redirect a launch
The launch goes to another signed-in account of the same provider. Model and conversation stay the same.
Block a launch
Stops launches of that provider before they spend quota: every launch, or only scheduled agents, webhook calls or manual launches.
Run a trigger
Starts one of your existing triggers, once per quota window. A cleanup brief when the week reaches 90%, for instance.
Move running agents
Switches the agents running on a saturated account to another signed-in account, same conversation. The account auto-switch is built on it.
Five templates to start from: Use another provider (Claude at 99% on any gauge, new agents on Codex), Use a lighter model, Use another account, Run a recovery trigger, and Alternate providers (Claude ×1 then Codex ×1).
Set it up in four steps
Footer usage badge, then the Quota management card of the Usage panel.
Sign in to Claude and Codex
Both CLIs, from the app. Their quota bars show up in the Usage panel of the footer, one per account.
Pick the Alternate providers template
In the Quota management card, it creates Claude ×1 then Codex ×1. The editor draws the rule as a diagram: the provider whose launches are shared, an arrow, and the rotation. Click a box to open its settings.
Shape the rotation
Choose Number of launches or Quota used, add or remove rows, set each share, account and model, then limit the rule to backlog tickets or to one project if you want.
Read the sentence, then save
Under the diagram the rule is written out in full, and the line Right now says what it would do with the current reading. Drag the rule in the list to set its priority.
Your agents can manage the rules too
Through the AgentsRoom MCP server, an agent can list the rules, create or edit one, reorder or delete it, and read every quota bar first. A rule written by an agent appears in the Quota management card like one you wrote, and the tools ask the agent to check with you before adding a rule that blocks or redirects launches.
quota_rules_listquota_rules_savequota_rules_deleteusage_overviewLimits, said plainly
The desktop app must be open: it reads the gauges and applies the rules. The phone edits the rules, but an agent launched directly from the phone does not go through the launch actions.
The position of the rotation, whose turn it is, is kept on the computer that runs the rule.
Only providers that publish a quota can be measured: Claude, Codex, Grok, GitHub Copilot CLI, Kimi, Cursor, Antigravity and OpenCode. Cursor and Antigravity have one gauge for the whole machine, so they have no second account to redirect to.
A rule reacts on the usage scan, so an action starts shortly after a threshold, not at the exact second. A threshold rule fires once per quota window, then re-arms at the reset.
A manual account pin on an agent always wins over a rule.
One question per page
Several pages talk about the same quota bars. This one answers how to spread the work over two subscriptions.
How do I use Claude and Codex together?
This page. A rotation shares new agents between providers or accounts, and threshold rules decide what happens near the limit.
My agent is already blocked mid-task, what now?
The account auto-switch: a quota rule moves the running agent to another account and the same conversation continues.
Account auto-switchCan I be warned before the limit?
Usage alerts: a desktop notification or a push on your phone at 50%, 75% and 90%, once per window.
Usage alertsHow do I run two Claude accounts on one computer?
Multiple Claude accounts: sign in to several, pin one per project, override it per agent.
Multiple Claude accountsFAQ
Can I use Claude Code and Codex together?
Yes. AgentsRoom runs both CLIs side by side in the same window, each agent in its own terminal. Since 6 October 2026 a quota rule can also decide for you which one a new agent starts on: one out of two on Codex, or a share of each quota in turn.
Can I alternate Claude Code and Codex automatically, one ticket out of two?
Yes, with the Alternate providers action in Number of launches mode. Claude ×1 then Codex ×1 sends one new agent out of two to Codex. Claude ×2 then Codex ×1 gives two for one.
How do I share my quota between Claude and Codex instead of burning one plan first?
Use the Quota used mode. Each row of the rotation spends its share of its own gauge, session or week, then hands over, and a gauge that resets ends the turn. For example 10% of the Claude session, then 10% of the Codex week. Both plans move at the pace you chose, so neither week resets unused.
My Codex plan has no 5-hour session, only a weekly limit. Does it still work?
Yes. A provider that publishes no session window is measured on its week, so a share of 10% means 10 points of the Codex weekly bar.
Can I combine two Claude accounts the same way?
Yes. Put Claude twice in the rotation and pick a different account on each row. The account field appears as soon as Claude has two signed-in accounts. In Quota used mode each row reads the gauge of its own account.
Can I limit it to the backlog tickets of one project?
Yes. Set Which launches to Backlog tickets and pick the project. Agents launched by hand, by a schedule or by a webhook then start as usual.
Will an agent switch from Claude to Codex in the middle of a task?
No. Only brand-new agents are shared out. An agent already running keeps its tool, and a restarted or resumed agent comes back on the provider the rotation gave it, with its conversation.
Which model does the agent use on Codex?
The one set on the row of the rotation. Agent's model keeps the model the agent would have started with, and Default model uses the default of the chosen CLI. You can also force a model and a reasoning effort per row.
What happens when Claude hits its usage limit during a rotation?
A provider at 99% or more skips its turn, so new agents go to the next row. If every row is full, launches go through unchanged. To also move the agents already running, add a rule with the action Move running agents.
Can I switch to Codex only when Claude is almost full?
Yes, that is the Use another provider template: when any Claude gauge reaches 99%, new agents start on Codex. With By source model, each Claude model gets its own target, pre-filled by capability tier: Fable to GPT-6 Astra, Opus to GPT-6 Sol, Sonnet to GPT-6 Luna.
Does it work with providers other than Claude and Codex?
Yes. Number of launches can share the launches of any provider you have enabled, since it counts launches. Quota used needs a provider that publishes a quota: Claude, Codex, Grok, GitHub Copilot CLI, Kimi, Cursor, Antigravity and OpenCode.
Two rules match the same launch. Which one wins?
The higher one in the list, whatever its action. A threshold rule placed above a rotation takes over when its condition holds. Drag a row to change the order.
Can I edit the rotation from my phone?
Yes. In the mobile app, Activity, Usage segment, Quota management opens the same rules, rotation included, and the desktop applies what the phone edits. An agent launched directly from the phone does not go through the launch actions.
Does it cost anything, and does AgentsRoom touch my credentials?
Quota rules are part of the desktop app. They use the subscriptions you already own and are signed in to: each CLI talks to its provider directly, AgentsRoom never pools anyone else's quota and never sees your credentials.
You may also like
Account auto-switch
The wall itself: an agent that runs out mid-task continues on another account you are signed in to, conversation included.
Usage alerts
A notification before the limit, at 50%, 75% and 90% of every quota bar. Same bars, same panel.
Multiple Claude accounts
Several Claude subscriptions on one computer, pinned per project or per agent. The accounts a rotation can alternate.
Multiple Codex accounts
Several ChatGPT accounts for Codex side by side, each with its own quota bars.
Multi-provider
Claude Code, Codex and a dozen other CLIs in one window, each agent on the tool you pick.
AI model equivalents
Which Codex model matches which Claude model, tier by tier. The table a rule uses to pick its targets.
Further reading
Claude Code vs Codex CLI: which one to pick in 2026 (and why not both)
An honest comparison of Claude Code and Codex CLI: pricing, speed, MCP, control. And why, after a few weeks, I stopped choosing between them.
How Many Tokens Do I Have Left in Claude? The Two Screens to Check.
Your Claude plan is not denominated in tokens, so no screen shows a token balance. What it actually meters, the two screens that do show it, why it drains while you are not typing, and what the September 2026 weekly-window change does to your week.
Claude Code Keeps One Login at a Time. Here Is How to Run Several.
A field guide to running a work account and a personal account on the same machine: the one environment variable that decides which login is active, why the shell approach breaks once you have more than two terminals, and how to pin an account per project.
Use both subscriptions you pay for
Download AgentsRoom, sign in to Claude and Codex, and pick the Alternate providers template. The next ticket knows where it goes.
Companion app: monitor your agents on the go
Bring your own: Claude, Codex, Antigravity CLI, or other AI provider.
Push bugs and requests straight to your public backlog.