llmcloud.ai
Migrate · OpenRouter

Two lines. Every model still works.

llmcloud.ai speaks the same OpenAI-compatible protocol as OpenRouter. Model IDs are compatible — provider/model names route to the same upstream — and you gain routing policies, semantic caching, guardrails and per-key IAM.

OpenRouter
import OpenAI from "openai";
const client = new OpenAI({
  baseURL: "https://openrouter.ai/api/v1",
  apiKey: process.env.OPENROUTER_API_KEY,
});
await client.chat.completions.create({
  model: "anthropic/claude-3.5-sonnet",
  messages: [{ role: "user", content: "Hi" }],
});
llmcloud.ai
import OpenAI from "openai";
const client = new OpenAI({
  baseURL: "https://api.llmcloud.ai/v1",   // ← 1
  apiKey: process.env.LLMCLOUD_API_KEY,     // ← 2
});
await client.chat.completions.create({
  model: "anthropic/claude-4.5-sonnet",     // same shape
  messages: [{ role: "user", content: "Hi" }],
});

What you gain

  • auto:quality / cost / speed / reasoning policies — no manual model juggling
  • Semantic caching with cache-hit routing (5–70% cost cut on repeat traffic)
  • Mid-stream failover across providers, not just retries
  • Per-key budgets, rate limits, region pinning, model allow-lists
  • PII redaction, prompt-injection classifier, content policies at the edge
  • Immutable audit log, SSO/SCIM, SOC 2 posture
  • MCP endpoint — mcp.llmcloud.ai works with Claude Code, Cursor, Windsurf
  • Transparent per-token billing, no ‘credits’ float, BYOK with zero markup
Are model IDs compatible?+

Yes — provider/model IDs (e.g. anthropic/claude-4.5-sonnet, google/gemini-3-pro) map 1:1. Where OpenRouter uses a version alias, we accept the same alias plus the fully-qualified ID.

Do OpenRouter's :nitro and :floor variants work?+

Substitute :nitro → auto:speed and :floor → auto:cost. The router picks a matching upstream per request instead of pinning a single one.

Can I dual-write during migration?+

Yes. Point staging at llmcloud.ai first, mirror requests with an X-LLMCloud-Shadow header, and compare cost + latency dashboards side-by-side before flipping prod.