One AI Gateway
for all your
The gateway is free forever. You pay the provider's list price, itemised on every receipt — we take nothing. Every provider is graded before it serves you.
$0 gateway fee · no prepaid credits · works with the OpenAI SDK
The gateway is free. Forever.
No routing cut, no credit float, no BYOK surcharge. You pay the provider's list price and see it itemised.
Every provider is graded.
A published 5-criterion rubric, our own probes, dated scorecards. Below the bar means out of default routing.
We name the model.
Per workload: one default, one cheap alternate, one escape hatch — and the directive that encodes it.
Routing, extended
Every request, routed with intent.
OpenRouter's one-endpoint promise, extended with the primitives agentic developers actually reach for.
auto:*
Smart routing
auto:quality, auto:cost, auto:speed, auto:reasoning — a contextual bandit picks the upstream that wins on the axis you care about, per request.
cache-hit routing
Semantic caching
Near-duplicate prompts get served from a shared cache instead of a model. Same answer, ~zero cost, sub-100ms TTFT.
thinking, unified
Reasoning trace
One field for step-by-step traces across o-series, Claude thinking, DeepSeek-R, and Qwen — no vendor-specific shim.
grounded
Web-grounded answers
Pass tools: ['web_search'] and the router picks the cheapest search-capable model. Citations returned in a standard shape.
one connector
MCP endpoint
Add mcp.llmcloud.ai to Claude Code, Cursor, or Windsurf and every model becomes a tool call — routed by the same policies.
multimodal in
Vision, audio, PDF
Send images, audio, and documents to any capable model through the same auto: policy. Router filters on modality automatically.
Opinionated defaults
We'll tell you which model to use
| Workload | Default pick | Cheap alternate | Directive |
|---|---|---|---|
| Coding agent | anthropic/claude-sonnet-4.5 | alibaba/qwen3-coder-480b | auto:coding |
| Function calling & agents | openai/gpt-5 | google/gemini-3.1-flash | auto:tools |
| Long-context retrieval | google/gemini-3.1-pro | google/gemini-3.1-flash | auto:long-context |
| Batch classification & extraction | openai/gpt-5.5-mini | openai/gpt-5.5-nano | auto:cost |
Platform
The gateway, operated.
Routing, caching, failover, governance and MCP — operated for you.
one endpoint
Unified API
Drop-in OpenAI-compatible base URL. Every model, every provider, one key.
94% → 99.9999%
Failover & reliability
Mid-stream failover across providers. Six-nines availability without lifting a finger.
keys · budgets · rbac
Governance built-in
Virtual keys with budgets, rate limits, region pinning. SSO, audit logs, SOC 2.
Migrate
From OpenRouter in two lines.
Same OpenAI-compatible protocol. Same provider/model IDs. Keep every tool call, streaming behavior and JSON-mode contract — gain routing policies, caching, guardrails and per-key IAM.
const client = new OpenAI({
- baseURL: "https://openrouter.ai/api/v1",
- apiKey: process.env.OPENROUTER_API_KEY,
+ baseURL: "https://api.llmcloud.ai/v1",
+ apiKey: process.env.LLMCLOUD_API_KEY,
});Trending
Top models this week
Get started
Live in three steps
Sign up
Create an account with Google, GitHub, or email. Spin up an org for your team any time.
Add a payment method
No prepaid credits, no minimums. You are billed the provider's list price in arrears — the gateway itself is $0.
Get your API key
Drop the key into the OpenAI SDK, point base_url at api.llmcloud.ai, and start streaming.
From the blog
Recent writing
Smart routing, explained: how auto:cost, auto:quality and auto:speed pick a model
A look under the hood of the llmcloud.ai router — feature extraction, live health scoring, and the bandit that decides which upstream wins each request.
Model ensembles: 12–18% accuracy gains by blending multiple LLMs
We ran a 40K-prompt eval across arithmetic, code, and long-form QA. Ensembling three mid-tier models beat GPT-4-class single-model calls at 30% lower cost.
What actually happens when OpenAI goes down
A post-mortem-style walkthrough of the June 2026 OpenAI incident from a gateway's perspective. How auto-failover kept 99.98% of routed traffic healthy.
Point your OpenAI SDK at us.
Pay the provider, not us. Swap one URL and get every frontier and open model at list price.