Beta · introductory pricing

Frontier models. One key. A tenth of the price.

TokenLowCost is a drop-in API endpoint for Claude, GPT, Grok and DeepSeek. Paste one base URL into Cursor, Claude Code or Codex, pick the model you want, and pay per token — prepaid, capped per key, billed to the cent.

$ curl https://api.tokenlowcost.com/v1/chat/completions
Works with
CursorClaude CodeCodexVS CodeClineRoo CodeZedOpenAI SDKAnthropic SDK
What you get

Infrastructure, not a wrapper.

A single endpoint that speaks every protocol your tools use, with the controls you'd expect from a real API provider.

One endpoint. Every protocol.

We detect the format from the request — Anthropic Messages, OpenAI Chat Completions or Responses — so the same base URL works everywhere. /v1 is optional.

POST/v1/chat/completionsOpenAI · Cursor · VS Code
POST/v1/messagesAnthropic · Claude Code
POST/v1/responsesCodex
GET/v1/modelscatalog

Budgets and expiry per key.

Create a key per project or teammate. Cap its spend, give it a lifetime, revoke it in one click. The gateway enforces it on every request.

Cursor — laptopsk-tlc-5f8a…e473
$1.24 of $2.00expires in 27d

Every request, priced.

Tokens in, tokens out, and the exact charge — per request, per key, per day.

modelclaude-opus-5.5
prompt1,204 tok
completion388 tok
charged$0.00041

Cache discounts, passed through.

Editors resend the same system prompt and files on every turn. That repeated prefix is billed at the cached rate — you get the discount, not us.

fresh input
1.00×
cached input
0.10×

The model you pick is the model you get.

No silent downgrades, no routing to a cheaper substitute. The response tells you which model answered.

requestedclaude-opus-5.5
servedclaude-opus-5.5match
Pricing

Per-token. Prepaid. About 90% below list.

Prices per 1M tokens. The crossed-out number is the model's standard list price for output. Pay only for what you use — no seats, no subscription.

Model Tier Context Input Output List (output) Saving
Loading prices…

Introductory pricing during beta. We'll give notice before rates change; your prepaid balance is never devalued.

Setup

Two lines. Keep everything else.

01

Create a key

Pick the model, set an optional budget and expiry. The key is shown once.

02

Point your tool at the base URL

Replace the provider URL with api.tokenlowcost.com and paste the key.

03

Keep working

Your editor, agent or SDK behaves exactly as before — only the bill changes.

Settings → Models

          

Full guides for every tool in the documentation.

Questions

Straight answers.

How can it be this cheap?
Two reasons. We're in beta and deliberately subsidizing early usage to grow — rates will rise as we exit beta, and you'll be told first. And editor workloads are cache-heavy: the same system prompt and files are resent every turn, and that repeated prefix is billed at the cached rate, which we pass straight through.
Do I really get the model I select?
Yes. Each key is pinned to the model you chose when you created it, and every response includes the model that actually answered. We never substitute a cheaper model.
Why is each key pinned to one model?
So the model name your tool sends doesn't matter — Cursor, Codex and others often send their own identifiers. You pick the model once in the dashboard; the key does the rest. Create as many keys as you need.
How does billing work?
Prepaid balance, deducted per request at the per-token rate of the key's model, to the fraction of a cent. Set a per-key budget to cap spend and an expiry to auto-disable. When the balance hits zero, requests stop — no overdraft.
How do I top up?
Self-serve payments are coming. During beta, balances are credited manually — ask us and we'll load your account.

Start with a $0 setup.

Create a key, paste the base URL, keep the tools you already use.