Gateway online · OpenAI / Anthropic / Gemini

One API key for your AI development workflow

Connect Claude Code, Codex, Gemini CLI, and image models through one developer account with unified usage, logs, and billing.

Create an account View pricing

Pay only for the API usage recorded in your account.

vibe-store / billing-engine Claude Code connected

src/billing/reconcile.ts usage.ts

41async function reserveUsage(accountId, cost) {
42  const account = await accounts.find(accountId);
43- if (account.balance < cost) return false;
44- account.balance -= cost;
43+ return db.transaction(async (tx) => {
44+   const row = await tx.lockAccount(accountId);
45+   if (row.balance < cost) return false;
46+   await tx.decrementBalance(accountId, cost);
47+   return true;
48+ });
49}

Claude Code streaming

You

Review the billing change and find concurrency risks before I merge it.

Claude

Potential race condition resolved The previous read-then-write flow could allow two requests to spend the same balance. Locking the account row inside db.transaction() makes the reservation atomic.

Suggested next step: add a concurrent reservation test for the exact remaining balance.

modelclaude-sonnet-4-6 latency1.28s cost$0.021

Designed for

Claude Code

Codex CLI

Gemini CLI

Image API

One gateway

Configure once. Keep your workflow.

Save one key. VibeLearning handles authentication, routing, usage records, and model access while responses stay inside the tools you already use.

vibe-store — Visual Studio Code VibeLearning connected

reconcile.ts

Gateway routeanthropic/messages Modelclaude-sonnet-4-6 Latency1.28s Cost$0.021

Protocols

Keep your tools. Change only the endpoint.

Claude Code, Codex, and Gemini CLI keep their native protocols. After setup, model selection, conversations, and tool use work as expected.

Interactive example — no live request is sent Open setup guide

ClientCodex CLI

Gatewayhalo.vibelearning.top

Endpoint/v1/responses

Request examplePOST

Example outcomeSimulated

Responses-compatible request

The gateway accepts the request through the OpenAI Responses path and records usage against the same account.

Protocol: OpenAI Responses
Endpoint: /v1/responses
Mode: Synchronous or streaming

Billing

Simple pay-as-you-go billing

Usage is charged against your account balance using the billing unit published for each model.

Main account

Pay as you go

Create an API key, choose an available model, and pay only for recorded usage.

Create account → Pricing

Ready to connect your first client?

Create an account, issue a dedicated API key, and verify the connection with one request.

Read the documentation