Skip to content
Operational · 99.9% uptime · ~40ms overhead
☕ Unified AI Gateway · v2.7

One endpoint for every model. One key, one balance.

Sign up, generate an API key, and route to Claude, GPT-5, and Gemini over one endpoint. Top up a USD balance once — CoffeeVibeAI meters usage, pools failover across upstream accounts, and updates your spend after every request.

No credit card required to sign up · pay-as-you-go USD balance

# one command, no manual config
curl -fsSL "https://api.coffeevibeai.com/install.sh?key=YOUR_KEY" | bash

  Verifying...
    OK  ~/.claude/settings.json
    OK  ~/.zshrc
    OK  No ANTHROPIC_API_KEY conflict

claude # writes shell env + ~/.claude/settings.json,
       # usage + balance updates after every request
# one endpoint, one sk- key
curl https://api.coffeevibeai.com/v1/messages \
  -H "Authorization: Bearer $CVAI_API_KEY" \
  -H "content-type: application/json" \
  -d '{"model":"claude-sonnet-5","messages":[...]}'

→ streaming response
event: content_block_delta
latency: 38ms (gateway overhead)
from anthropic import Anthropic

client = Anthropic(
    base_url="https://api.coffeevibeai.com",
    api_key=os.environ["CVAI_API_KEY"],
)

resp = client.messages.create(
    model="claude-sonnet-5",
    messages=[{"role": "user", "content": "ship it"}],
)
# same key works for GPT-5 and Gemini too
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.coffeevibeai.com/v1",
  apiKey: process.env.CVAI_API_KEY,
});

const res = await client.chat.completions.create({
  model: "gpt-5.4",
  messages: [{ role: "user", content: "ship it" }],
});
# .cursorrules → model routing
provider: coffeevibeai-gateway
base_url: "https://api.coffeevibeai.com"
auth: env:CVAI_API_KEY

# Cursor now reports usage per-project in the dashboard

Works with your stack

Native, out of the box.

Cursor Windsurf Claude Code VS Code / Cline Aider LangChain LlamaIndex Dify OpenWebUI NextChat
Core engine

Everything a gateway should do, nothing it shouldn't.

One account, one key, one balance — we handle the operational layer underneath.

100% Drop-in Compatible

Swap only the base_url and key. Same Anthropic and OpenAI SDKs, zero refactor.

Sub-50ms Relay Overhead

Rust-backed proxy layer adds negligible latency on top of the upstream provider's own response time.

Pooled Capacity & Failover

Behind your key, we pool multiple upstream accounts per provider. If one is rate-limited, requests fail over to another — automatically.

Real-Time Usage & Spend Analytics

Per-key token counts, cost breakdowns by model, and quota alerts — all metered against one balance, in one dashboard.

Zero Prompt Retention

Requests are proxied in memory. Prompts and completions are never written to disk on our side.

Built for two kinds of teams

Solo builder or shared workspace — same gateway.

For Developers

Indie hackers and AI engineers wiring up agents and apps.

  • One Key, Every ModelCall Claude, GPT-5, and Gemini with the same credential.
  • Pay-as-You-Go BalanceTop up once, spend it down across any model — no separate subscriptions.
  • 30-Second SetupOne environment variable, no SDK changes.
  • Quota AlertsGet notified before a runaway agent burns through your balance.

For Agencies & Teams — Reseller

Studios and platform teams managing keys for many clients at once.

  • Multiple Keys, One AccountIssue a separate sk- key per client or project, each with its own quota and rate limits.
  • Seller DashboardRoll up usage, cost, and per-model breakdowns across every customer account assigned to you.
  • Self-Service Key LookupShare a lookup link so a customer can check their own balance and usage, and run the Claude Code installer, without you in the loop.
  • Your Own PlansSellers with plan-authoring enabled can create and manage subscription plans scoped to their own customers.
Estimate your spend

One balance, published per-model rates.

Rates are listed up front in your dashboard before you spend a token — here's what a typical month looks like at these prices.

Model
Estimated monthly volume 50M tokens
Anthropic (Direct API) $225.00
CoffeeVibeAI balance spend80% less $45.00
ModelInput / MTokOutput / MTok
Claude Fable 5$10.00$50.00
Claude Opus 5$5.00$25.00
Claude Opus 4.8$5.00$25.00
Claude Sonnet 5$2.00$10.00
Claude Haiku 4.5$1.00$5.00

Rates above are Anthropic's direct API list price. CoffeeVibeAI bills your balance at 1/5th these rates.

Three steps

From signup to first request in under a minute.

01

Sign up & get a key

Create an account and generate an sk- API key from your dashboard.

02

Set one endpoint

Point your SDK or CLI at the gateway. Claude Code CLI users can skip the manual setup entirely with the one-line installer from the dashboard.

curl -fsSL "https://api.coffeevibeai.com/install.sh?key=sk-..." | bash
03

Top up & build

Add balance, then every request shows up in your dashboard — cost, latency, and model, in real time.

Frequently asked

Good to know before you get a key.

You top up a balance in your dashboard, and every request is metered in USD against that balance at published per-model rates. There's no separate provider subscription to manage.

Requests are proxied in memory and never written to disk. We recommend reviewing our zero-retention policy in the docs before routing sensitive workloads.

Yes. Both read the standard base URL environment variable, so switching is a one-line change with no plugin required.

We pool capacity across multiple upstream accounts per provider, so a single degraded account fails over automatically. If a whole provider is down, requests to its models return that provider's own error and you can switch models manually.

Yes. Create as many sk- keys as you need, each with its own optional quota and rate limits, so teammates or projects don't have to share one credential.

Ready to see all your AI spend in one place?

Get an API key, top up a balance, and start calling Claude, GPT-5, and Gemini in minutes.