VALUE
Codex limits, cheaper
$10 — Codex limits at half the price. $24 — 3× the limits for the same money.
- Plus: Codex limits at half price
- Pro: 3× the limits
- Weekly quota reset
Works with your tools
OpenClaw
OpenCode
Cursor
VS Code
Continue
Codex CLI
Cline
OpenAI SDK001What you get
Everything for coding, one subscription.
SuperCodex, honest limits, and a spend dashboard — in one account.
SUPERCODEX
SuperCodex — Codex, only stronger
Agent coding sessions in your IDE and CLI on Luma Cloud's capacity — the same limits, cheaper.
- The real GPT-5.5 xhigh in IDE and CLI
- Long agent sessions
- Like Codex — only better value
Dashboard
Plan, spend, and keys on one screen
See what is spent and what remains. No surprises at the end of the month.
- Real-time spend
- Limits and keys in one click
- Cancel right in the dashboard
Need your own API for CLI agents? That's Builder →
More work per dollar.
The subscription covers a full working day, and builder traffic costs up to 10x less than a direct API. The numbers below come from public price lists.
Headroom advantage
More working headroom on the same budget.
The same working budget inside Luma Cloud carries more continuous AI work, longer agent cycles, and longer sessions.
API advantage
API that costs less per request.
For integrations, background jobs, and apps, what matters is predictable product-traffic cost instead of a raw API path by default.
Typical direct API
A typical direct GPT-5.5 API per 1M.
Luma Cloud API
The same product traffic on Luma Cloud builder pricing.
Average mix assumes a 50/50 split between input and output.
003Everyday reality
Familiar pains — solved.
The painHow Luma Cloud solves it
Quota dies in the middle of a taskA full day of working headroom, reset weekly
Metered API bills feel scaryFlat subscription from $10/mo — spend visible instantly
Every tool needs its own setupOne account for the agent, IDE, CLI, and API
Assistant memory lives in someone else's cloudMemory and context stay on your machine
004Privacy
Your agent memory stays yours.
- Local-first mode keeps memory on your computer
- Hosted mode runs the agent on your dedicated server
- Luma Cloud is the access and power layer, not the owner of your data
005Pricing
Subscription for people.
API separate.
There are three SuperCodex subscriptions: Plus, Pro, and Builder. API is a separate tariff for product traffic, automations, and external agent systems.
Plus plan
Plus
SuperCodex at half the price of Codex — the same limits, the same real GPT-5.5 xhigh, just $10.
$10
/month
Early-access price
Price locked for as long as you stay subscribed.
$10 charged every month
What you get
SuperCodex in your IDE and CLI — on the real GPT-5.5 xhigh
Codex limits at half the price
Spend dashboard and weekly quota reset
Most popular
Pro plan
Pro
3× the limits of Codex at $20 — for heavy days and long sessions.
$24
/month
Early-access price
Price locked for as long as you stay subscribed.
$24 charged every month
What you get
3× the limits of regular Codex
The same real GPT-5.5 xhigh, same dashboard and install path
Weekly quota reset, spend visible from the first request
Higher ceiling for one heavy user without selling it as literal unlimited.
API keys included
Builder plan
Builder
SuperCodex 5x plus your own keys for CLI and external agents; top-ups and API-only usage sit on top.
$49
/month
Early-access price
Price locked for as long as you stay subscribed.
$49 charged every month
What you get
5× the Plus limits for SuperCodex
Up to 5 API keys for any CLI or external agent
Keys burn Builder quota; when it runs out, use top-ups or metered API
Abuse guardrail: keys are not unlimited, and spend is visible in the dashboard.
Fair use applies: if the system sees abuse, account sharing, or multi-user load behind one key, speed and throughput may be limited.
API tariff
API is paid by tokens, not by subscription.
For products, CLIs, and external agents: keys, ledger, and billing by actual usage.
Direct API price
Input$0.50 / 1M
Output$3.00 / 1M
Average cost follows your input/output mix, without an invented fixed rate.
OpenAI-compatible endpoint
https://api.lumaos.cloud/v1gpt-5.5-xhighgpt-5.5-highgpt-5.5gpt-5.5-mediumgpt-5.5-low
One GPT-5.5 line — including the real gpt-5.5-xhigh. The same canonical ids that /v1/models returns.
How to connect
Grab a key in the dashboard — any OpenAI-compatible client works. Full guide in the docs.
006Start in 10 minutes
Four steps — and the agent is working.
Sign in
An account takes a minute — no card required to look around.
Pick a plan
Plus to start, Pro for heavy days. Cancel anytime.
Install SuperCodex
Download, open, sign in. No Node.js, no terminal.
Work
Spend and remaining quota are visible from the first request.
From payment to the first agent session usually takes under 10 minutes.
007FAQ
Common questions.
In the dashboard, one click. Access stays until the end of the paid period — no emails, no retention tricks.
Email us in the first few days and we'll sort it out, person to person. The formal terms live in the Refund Policy below.
No. You sign up with an email code or Google — no card. A card is only needed at the Stripe payment step, when you pick a plan.
Yes. We do not swap models: requests run on the same GPT-5.5 with the xhigh reasoning preset as official Codex — 100% the real thing. Canonical model ids are visible in /v1/models, and spend is visible in your dashboard from the first request.
Yes. Any OpenAI-compatible client — IDE, CLI, or SDK — connects to a single address. SuperCodex installs on a Mac in minutes.
Plus covers a full working day of agent sessions. Pro is roughly three times the Plus quota — for heavy days and research loops. The quota resets weekly and the remainder is always visible in the dashboard.
Plus, Pro, and Builder cover SuperCodex. API access stays separate and turns on by request, so billing stays readable and honest.
Yes. Once apps, automations, or shared workflows appear, keys, caps, and revoke actions stay in the same account but open as a separate builder route.
No. The normal path starts in the dashboard and SuperCodex. Builder keys are requested later only when product traffic, automations, or external tools appear.
Persistent memory stays inside your own agent base. In local-first mode it lives on your computer, and in hosted mode it lives on your dedicated server while Luma Cloud remains the access and compute layer.
Still have a question? Write to us — we answer fast. Contact us →
