BetaQuota is live — one prepaid balance for Cursor, OpenCode, Zed and every agent.
Quota
Start free
Infrastructure layer

One green gateway for every coding agent.

Buy Quota Tokens once, spend them across model routes, and keep Cursor, OpenCode, Zed and autonomous agents on one calm, controlled API surface.

Start freeRead the docs
1 balance
for every coding agent
6+ routes
fast → opus, one API
0 raw prompts
stored by default

Works with the tools you already ship in

OpenAI-compatibleCursorOpenCodeZedJetBrainsClineAPI / SDKCodexOpenAI-compatibleCursorOpenCodeZedJetBrainsClineAPI / SDKCodex
Solutions

Premium infrastructure, not a magic wrapper.

Three pillars keep Quota boring on purpose: one surface, transparent cost, and usage you can actually trust.

One unified gateway

Point Cursor, OpenCode, Zed, JetBrains or your own agent at a single OpenAI-compatible endpoint. Swap models with a route alias — never touch client code again.

Cost-aware routing

Each route carries a transparent multiplier, so an agent only spends what a task deserves. Cheap routes for routine work, premium reasoning only when it matters.

Usage you can audit

Every request returns normalized usage. Watch tokens burn down by the hour and by model, set per-key limits, and keep experiments isolated from production.

The dashboard

See exactly where every token goes.

A live ledger of tokens remaining, requests served and spend you saved — plus an hourly breakdown of which model burned what. No black boxes.

  • Hourly usage by model, streamed live
  • Per-key limits to isolate agents and experiments
  • Savings vs. paying each provider directly
Balance
8.42M tokens
Live
Requests
12,183
Saved
$1,240
Avg ×
2.4
Tokens burned · last 12hquota/sonnet
Route matrix

Pick the price you pay, per request.

Every model maps to a route alias with a clear multiplier. Cost is visible before the agent spends a single token.

RouteMultiplierStatus
quota/yunyi-gpt-5-4
×1
Operational
quota/yunyi-gpt-5-5
×1
Operational
quota/codex-gpt-5-6-sol
×2
Operational
quota/codex-gpt-5-6-luna
×2
Operational
quota/codex-gpt-5-6-terra
×2
Operational
quota/codex-gpt-5-5
×1
Operational
Developer path

One key. One balance. Clean streaming.

Standard request shape, stable route aliases and a streaming lifecycle that ends cleanly — content, final chunk, [DONE], close.

Integration guides
bash
curl https://api.quota.host/v1/chat/completions \
  -H "Authorization: Bearer qta_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "quota/sonnet",
    "messages": [{ "role": "user", "content": "Review this diff" }],
    "stream": true
  }'
Trust layer

Numbers with a source. Routes with a state.

Quota should feel like infrastructure. The interface shows where usage came from, which route handled it, and exactly why a balance changed.

99.9%
route availability target
<1.8%
normalized error rate
0
raw prompts stored
6+
models, one endpoint
Get started

Put a budget gate in front of your agents.

Start with one key, one balance, one route. Scale only after the usage is visible.

Start freeView pricing
FAQ

Answers before Quota sits in front of your tools.

What are Quota Tokens?

Prepaid usage credits. Requests consume tokens by the route multiplier you choose, so cost is predictable before an agent starts spending.

Does Quota store prompts?

No raw prompts are stored by default. Only normalized usage metadata is kept for billing, diagnostics and request logs.

Does it work with Cursor & OpenCode?

Yes. Add Quota as a custom OpenAI-compatible provider, drop in your key, and select a Quota route alias. Guides walk you through every IDE.

Why do routes have multipliers?

Models differ in cost and capability. Multipliers make that visible up front instead of surfacing as a surprise on the invoice.

How do I pay?

Top up a balance with a slider — the more you load, the cheaper each token gets. Payment is in crypto today.

What if a route is unavailable?

Quota returns a normalized error and leaves your balance untouched for failed upstream requests.