Router One

A managed alternative to a self-hosted LiteLLM proxy

LiteLLM is a popular open-source proxy you deploy and operate yourself. Router One is a fully managed, OpenAI-compatible LLM API gateway that covers the same routing and fallback needs — plus built-in billing, observability, and worldwide reachability including Mainland China — with no infrastructure to run. Use it when you want one endpoint for 40+ supported models such as GPT, Claude, Gemini, and DeepSeek, plus Claude Code and Codex CLI workflows.

Managed gateway vs self-hosted proxy

FeatureRouter OneLiteLLM (self-hosted)
Hosting modelFully managed — nothing to deploy or operateOpen-source proxy you self-host (managed tier also available)
Smart routing & fallbackEWMA latency + posted cost + 5xx rate, managedConfigurable routing and fallback you maintain
ObservabilityPer-request traces, tokens, and cost out of the boxLogging/observability you wire up (e.g. Langfuse)
OpenAI-compatible APIYes — Chat Completions, drop-in base URLYes — OpenAI-compatible proxy
Provider accessUpstream access included through one Router One keyYou supply and manage each provider API key
Accessible from Mainland ChinaChina-reachable endpoint with published latency benchmarkDepends on where you host the proxy
Billing & paymentBuilt-in wallet: Stripe, USDT/USDC, WeChat Pay, AlipayBring your own provider keys and billing

Where a managed gateway helps

No proxy to run

Self-hosting LiteLLM means provisioning, scaling, patching, and monitoring a proxy. Router One is managed, so you skip the ops and start with an API key.

Billing built in

Top up by card through Stripe, with USDT and USDC on six chains, or in RMB with WeChat Pay and Alipay — one wallet, no separate provider invoices to reconcile.

Reachable worldwide, including Mainland China

Hosting an open-source proxy does not solve reachability for a distributed team on its own. Router One's managed endpoint serves members everywhere, including Mainland China. Based on the China latency benchmark last updated 2026-05-15, Router One measured 110-130ms p50 across Beijing, Shanghai, and Shenzhen; individual networks may vary.

View the China latency benchmark ->

Switch with one base URL change

Both Router One and a LiteLLM proxy speak the OpenAI Chat Completions API, so moving over is usually a base URL and key change. Point your client at Router One and drop the keys you no longer need to manage.

terminal
# Replace your LiteLLM proxy base URL with Router One
export OPENAI_BASE_URL=https://api.router.one/v1
export OPENAI_API_KEY=sk-your-router-one-key

FAQ

Is LiteLLM open source?

Yes. LiteLLM is an open-source, self-hostable proxy, and it also offers a managed tier. Router One is a managed gateway — the trade-off is run-it-yourself flexibility versus zero-ops managed access with billing and observability built in.

Can I migrate from a LiteLLM proxy?

Yes. Both are OpenAI-compatible, so most integrations move with a base URL and API key change. Tools and SDKs that work against a LiteLLM proxy work against Router One.

Do I still bring my own provider keys?

No. Router One includes upstream access and billing through a single key and wallet, so you do not manage a separate key per provider. One endpoint reaches 40+ supported models — browse them at /models.

Which should I choose?

If you want full control and are happy to self-host, LiteLLM is a strong open-source option. If you want a zero-ops gateway with upstream access, billing, and observability built in — reachable worldwide, including Mainland China — Router One removes the infrastructure work.

More comparisons

Want a managed gateway instead of running a proxy?

Get your API key