workspace · 5 accounts · synced just now · local-only

your codex account pool, in one local console

inventory every Codex identity on this machine · one-tap smart switch · unified local proxy routing · everything stays on this device.

accounts

05

3 active · 1 idle · 1 rate-limited · 0 exhausted

pool quota · 7d

62%

used 128.4k / 207k tokens

cost 7d · local est.

$42.07

▲ +12.4% vs prev 7d

active identity

brian.dev

switched 2h ago · auto-selected

5h quota 64% · 7d quota 52%

0 selected tick the rows you want to export

accounts

click any account to expand details · click switch to set active

// token usage

24h23.53M

3d26.21M

7d49.46M

30d1.6B

▸ how it works

strategy chatgpt-first → highest-7d-remaining · last rotation 2h ago: team-research hit 5h limit → fallback to brian.dev

next pick · brian.dev · then design-night

▸ account pool · 5 / capacity active 1 rate-limited 1 idle 1 reauth 1 disabled 1

01

brian.dev brian@studio.io

pro · chatgptoauth · since 2025-12-04

5h64%

7d52%

current

usage overview

5h usage64%

remaining 36%5h

7d usage52%

remaining 48%1w

account meta

reset time2026/06/17 21:49

planPRO

recent switches

switched to this account2026/06/17 16:57brian@studio.io

quick actions

02

team-research research@studio.io

pro · chatgptoauth · session imported

5h96%

7d78%

rate-limited

usage overview

5h usage96%

remaining 4%5h

7d usage78%

remaining 22%1w

account meta

reset time2026/06/17 19:30

planPRO

recent switches

fell back to brian.dev (5h limit)2026/06/17 14:42auto

quick actions

03

design-night design@studio.io

plus · chatgptoauth · since 2026-01-12

5h31%

7d42%

idle

usage overview

5h usage31%

remaining 69%5h

7d usage42%

remaining 58%1w

account meta

reset time2026/06/18 02:10

planPLUS

recent switches

no recent switches.

quick actions

04

qa-bot qa@studio.io

api · relayapi key · proxied

5h12%

7d18%

reauth required

usage overview

5h usage12%

remaining 88%5h

7d usage18%

remaining 82%1w

account meta

reset time—

planAPI

recent switches

refresh_token revoked · re-authorize required.

quick actions

05

archive-2025 legacy@studio.io

free · chatgptimported · auth.json

5h4%

7d8%

disabled

usage overview

5h usage4%

remaining 96%5h

7d usage8%

remaining 92%1w

account meta

reset time—

planFREE

recent switches

account disabled · enable to use.

quick actions

add account

usage analytics

cost & usage analytics

local estimate from your Codex session logs · token consumption, cost by project, hourly heatmap, and top sessions — never leaves this machine.

exporting…

64%

total cost

$1,284.07

all time · local est.

cost · last 7d

$42.07

▲ +12.4% vs prev 7d

total tokens

1.62B

in 23.4M · out 1.59B

sessions

3,418

128 projects

weekly budget

tracked against this week's local cost estimate

this week$42.07 / $80.00

remaining $37.93 · resets monday53%

cost by project · top 10

studio-web

~/dev/studio-web

842 sessions · 6.1k prompts · 18k events

$312.40

codex-tools

~/dev/codex-tools

561 sessions · 4.0k prompts · 12k events

$231.10

data-pipeline

~/work/data-pipeline

402 sessions · 3.1k prompts · 9k events

$181.55

mobile-app

~/dev/mobile-app

288 sessions · 2.2k prompts · 6k events

$128.90

infra-scripts

~/ops/infra-scripts

176 sessions · 1.4k prompts · 4k events

$84.20

activity heatmap · hour × weekday

recent sessions · top 80

session	project	model	tokens	cost	updated
sess_8f2a	studio-web	gpt-5-codex	4.2M	$8.41	2026/06/17 16:57
sess_71bd	codex-tools	gpt-5-codex	2.8M	$5.62	2026/06/17 14:12
sess_5c09	data-pipeline	gpt-5	1.9M	$3.80	2026/06/17 11:48
sess_3e44	mobile-app	gpt-5-mini	920k	$1.10	2026/06/16 22:30
sess_2a18	infra-scripts	gpt-5-codex	640k	$1.28	2026/06/16 19:05

top prompts by cost

01refactor the entire authentication module to use the new session token rotation strategy…$4.20

02generate comprehensive integration tests for the proxy load-balancer covering all edge cases…$3.15

03migrate the data pipeline from the legacy schema to the partitioned table layout…$2.88

updated · 2026/06/17 16:58 source files · 3,418 failed sources · 0 pricing source · bundled table

local api proxy

local api proxy & routing

expose your account pool as a single OpenAI-compatible endpoint · load-balance across identities, manage proxy keys, deploy to remote servers, and tunnel via cloudflared.

status

● running

uptime 4h 12m

base url

127.0.0.1:8788

openai-compatible · /v1

pool size

5 accounts

3 active · 1 idle · 1 rate-limited

proxy key manager

sk-proxy-••••••••••••••3f9a

label	key	requests	created
default	sk-proxy-…3f9a	12,408	2026-05-01
ci-runner	sk-proxy-…7b2c	3,991	2026-05-18

load balance strategy

sequentialuse one account until it nears the limit, then move to the next.

round-robinrotate evenly across all healthy accounts per request.

5h limit threshold%

supported models

gpt-5-codex gpt-5 gpt-5-mini gpt-4.1 o3

remote servers

prod-relay-01

root@10.0.4.21 · :8788

eu-edge-02

deploy@10.0.7.8 · :8788

[2026-06-17 16:40:02] proxy listening on 0.0.0.0:8788 [2026-06-17 16:40:02] pool: 5 accounts loaded · 3 healthy [2026-06-17 16:41:18] GET /v1/models 200 · 12ms

cloudflared tunnel

https://relay.studio.trycloudflare.com

reference radar

github inspection results

scan GitHub for projects related to local Codex account management · borrow ideas, watch licenses. runs manually by default; logs stay in the browser.

current default

manual refresh

scans run only when you press refresh.

optional capability

daily auto-scan

opt-in scheduled scan once per day.

log location

browser-local logs

results cached in local storage only.

candidates

18

high relevance

4

license types

MIT · Apache

last refresh

2h ago

acme/codex-pool-managerhigh

Desktop manager for rotating multiple Codex/ChatGPT identities with a local OpenAI-compatible proxy and per-account quota tracking.

★ 2.4k⑂ 188! 23TypeScriptMITupdated 3d ago

reusable: sequential-then-round-robin balancer design · per-key request metering.

caution: bundles a telemetry client — strip before reuse.

opensrc/auth-rotatormedium

CLI tool that rotates auth.json files for several AI providers and keeps session tokens warm.

★ 640⑂ 47! 9RustApache-2.0updated 12d ago

reusable: token keep-alive scheduler approach.

local rules

how it works · principles

the design rules behind account rotation, the local proxy, inspection, and analytics — and the boundaries each one respects to keep everything local and safe.

default boundary

local-first

no credential or usage data leaves this device unless you explicitly export it.

inspection mode

manual trigger

github radar never runs on its own — you press refresh.

analytics basis

local estimate

cost and tokens are estimated from local session logs against a bundled price table.

GitHub Radar

inspection logic

how reference projects are discovered, ranked, and surfaced without ever auto-running.

working rules

queries github only when you press refresh.
ranks candidates by name, description, and topic match.
caches results locally so repeat views cost no calls.

guardrails

no credentials are ever sent to github.
respects github rate limits with backoff.

Local API Proxy

proxy logic

how the pool is exposed as one endpoint and how requests are routed across identities.

working rules

binds to localhost and speaks the OpenAI-compatible api.
routes by the selected strategy (sequential or round-robin).
skips rate-limited or disabled accounts automatically.

guardrails

proxy keys are stored locally and shown masked.
remote deploy requires your explicit action.

Account Rotation

rotation & smart switch

how the active identity is chosen and when an automatic fallback happens.

working rules

prefers chatgpt accounts, then highest 7d remaining.
falls back when the current account hits its 5h limit.
manual switch always overrides automatic choice.

guardrails

disabled accounts are never auto-selected.
switching never deletes or overwrites stored auth.

Session Safety

login state rules

how authenticated sessions are kept valid and isolated from each other.

working rules

each identity keeps its own auth.json snapshot.
switching swaps the active auth atomically.
expired tokens are flagged for re-authorization.

guardrails

tokens are never logged in plaintext.
export is redacted unless you opt into full export.

Usage Analytics

analytics basis

how cost and token figures are produced and why they are estimates.

working rules

parses local session logs only.
applies a bundled price table per model.
aggregates by project, hour, and session.

guardrails

figures are estimates, not official billing.
no log content is uploaded anywhere.

preferences

settings

language, appearance, launch behavior, and editor integration — all stored locally on this device.

general

language

interface language for this app.

theme

light or dark appearance.

tray usage display

what the menu-bar icon shows.

launch & integration

launch at startup

open this manager automatically when you log in.

launch codex after switch

start Codex automatically right after switching account.

launch codex as admin

request elevated permissions when starting Codex.

smart switch includes api accounts

allow api-key accounts to be picked by smart switch.

codex launch path

custom executable used to start Codex.

/usr/local/bin/codex

sync opencode openai auth

mirror the active auth into opencode on switch.

restart opencode desktop on switch

relaunch opencode desktop so it picks up the new auth.

restart editors on switch

relaunch selected editors after switching account.

editor targets

which editors to restart.