Ship in five minutes.

curl https://api.llmcloud.ai/v1/chat/completions \
  -H "Authorization: Bearer $LLMCLOUD_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto:quality",
    "messages": [{"role":"user","content":"Draft a launch tweet."}]
  }'

Pass a routing directive as the model name. We score every eligible upstream on capability, live latency, price, and quality benchmarks — then pick the winner.

import OpenAI from "openai";
const client = new OpenAI({
  baseURL: "https://api.llmcloud.ai/v1",
  apiKey: process.env.LLMCLOUD_KEY,
});

const stream = await client.chat.completions.create({
  model: "auto:quality",
  stream: true,
  tools: [{ type: "function", function: { name: "getWeather", parameters: {} } }],
  messages: [{ role: "user", content: "Weather in Tokyo?" }],
});
for await (const chunk of stream) process.stdout.write(chunk.choices[0]?.delta?.content ?? "");

Every request carries a bearer key on the Authorization header. Keys are scoped objects, not bare strings: each one carries an environment label, an optional monthly spend ceiling, an allow-list of models or routing policies, and an optional provider allow-list. A key that exceeds its ceiling returns 402 budget_exceeded rather than silently continuing to spend.

# Environment-scoped keys
export LLMCLOUD_KEY="lc_live_..."   # production, budget capped
export LLMCLOUD_KEY="lc_test_..."   # staging, routed to cheap models only

curl https://api.llmcloud.ai/v1/models \
  -H "Authorization: Bearer $LLMCLOUD_KEY"

Bringing your own provider contracts? See BYOK — your upstream keys are used for billing while routing, caching, and analytics still apply.

Every completion returns the full accounting of what happened, so you never have to guess which upstream served a request or what it cost.

Errors use OpenAI-compatible shapes so existing client error handling keeps working. The gateway already retried across upstreams before returning anything in the 5xx range — an error that reaches you means every eligible candidate failed.

Every request returns x-llmcloud-route and x-llmcloud-cost. See the usage dashboard for spend, latency, and per-model traces.

curl "https://api.llmcloud.ai/v1/usage?from=2026-08-01&to=2026-08-31&group_by=model" \
  -H "Authorization: Bearer $LLMCLOUD_KEY"

{ "rows": [
    { "model": "llama-4-maverick", "provider": "fireworks",
      "requests": 41230, "input_tokens": 88.2e6, "output_tokens": 12.4e6,
      "cost_usd": 214.87, "platform_fee_usd": 0.00, "cache_hit_rate": 0.31 }
] }