Accounts Pool: shared quota and load balancing for saved Codex accounts
Accounts Pool is the CodexUse Pro feature that turns selected saved Codex accounts into one local API, so another app gets shared quota, load balancing, and failover instead of depending on one account at a time.
Why this matters
Before Accounts Pool, another app only had whatever headroom one account gave it. If that account was close to quota or hit a runtime issue, the workflow stalled fast.
Accounts Pool changes that. Another app gets one local API and one pool key, while CodexUse pools the selected accounts behind it so fresh work can be load-balanced and blocked sessions can roll over.
What shipped
- One local OpenAI-compatible API for the saved accounts you select
- Pool keys for the apps you connect to it
- Shared quota across selected saved accounts
- Load balancing for fresh requests across the pool
- Published model names for downstream apps
- Failover when an eligible account runs out or fails
- CLI commands to manage the pool and inspect a headless daemon runtime
Where it fits
| Use case | Why Accounts Pool helps |
|---|---|
| Local agent runner | Give it more total headroom by pooling selected accounts behind one local API. |
| Sidecar editor extension | Keep the editor integration simple with one key, while CodexUse load-balances new work in the background. |
| Longer response workflows | When one account hits quota or a runtime issue, the pool can roll to another eligible account instead of dying immediately. |
What it does not pretend to solve
- It is not a public hosted proxy. The point is a local endpoint on your own machine.
- It does not create extra quota from duplicate accounts. If two saved profiles are really the same underlying OpenAI account, they still share the same headroom.
- It does not hide bad requests. Unsupported or malformed requests still fail instead of being silently retried somewhere else.
What users actually get
You save Codex accounts in CodexUse once. Then another app on your machine gets one API, one key, more shared quota, load balancing across selected accounts, and automatic rollover when one account gets blocked.
That makes Accounts Pool useful anywhere you already have an OpenAI-compatible app or script and want CodexUse to be the pooled account layer behind it.
Getting started
- Open Settings → Accounts Pool in CodexUse.
- Enable the pool and choose the profiles you want available.
- Publish the model names you want the other app to see.
- Generate a local pool key.
- Point the other app at the base URL shown by CodexUse.
CLI and daemon path
If you want the same flow without opening the desktop app, the CLI now covers the operational side too:
codexuse account-pool status
codexuse daemon start --port=3773
codexuse account-pool status --runtime=daemon --port=3773
codexuse account-pool keys create --runtime=daemonThe settings stay shared with the desktop app. Keys and pooled sessions stay runtime-specific, so a daemon can keep its own pool store and one stable local base URL. The daemon requires Bun on PATH, and a daemon URL needs a daemon-scoped key created with --runtime=daemon.
Related
Is Accounts Pool part of Free?
No. Accounts Pool is a Pro feature.
Does it use my normal Codex login?
Yes. CodexUse still uses the same Codex CLI OAuth profiles you already saved locally.
Do downstream clients need my OpenAI credentials?
No. They use a local pool key generated by CodexUse, not your Codex account credentials.