Curated models for your best work

Route every model. Read every receipt.

One OpenAI-compatible endpoint for 87 models. No markup on tokens and an exact receipt for every request.

New Gemini 4 Argon from Google is now routed. $2 in, $10 out per 1M tokens.

TokenSpend account required. Routing on your own keys is free.Get API key

87: models
18: vendors
2.1B: tokens routed

Routes every frontier model

OpenAIAnthropicAmazonxAIMetaGoogleZ.aiDeepSeekAlibabaThinking MachinesMoonshotMiniMaxMistralNVIDIAXiaomiStepFunTencentinclusionAI

Trusted by engineers at

Start in three steps

1Sign in
Create an account with GitHub or email. You can set up a team later.
2Add a key or credits
Top up credits, or connect a provider key you already have.
3Paste the base URL
Point any OpenAI-compatible client at one base URL with your TokenSpend key.

Featured models

Price per 1M tokens and context window

View all 87 models

Claude Opus 5.5Anthropic

$4.00 in · $20.00 out1M context

−60%

GPT-6 LunaOpenAI

$0.10 in · $0.50 out1.05M context

+3200%

Gemini 3.8 FlashGoogle

$0.75 in · $3.75 out1.05M context

−40%

Grok 4.7xAI

$2.00 in · $6.00 out500K context

−26%

Kimi K3Moonshot

$3.00 in · $15.00 out1.05M context

−60%

GLM-5.3Z.ai

$0.90 in · $3.00 out1.05M context

−68%

Everything you need to build with AI

One endpoint, exact receipts, and spend joined to merged PRs.

POST /v1/chat/completions
{
  "model": "claude-sonnet-5",
  "messages": [{"role": "user"}]}
}

One endpoint

Drop in any OpenAI SDK and switch models by name.

Request completed$0.0024

Modelclaude-sonnet-5

Input tokens1,024

Output tokens256

Total tokens1,280

Exact receipts

Every request returns the model, token counts, and exact USD.

Shipped 61%In flight 24%Did not ship 15%

PR 128 Checkout retryShipped

PR 131 Dashboard chartsIn flight

PR 126 Vector searchDid not ship

Spend that shipped

Join coding-agent spend to the pull requests that merged.

Start routing. Keep every receipt.

The router is free and providers are billed at list price.

Get API key Read the docs