Smart routing gateway for China's top providers

One API for China's Top Models
Top providers. Resilient access. No markup.

Our own gateway keeps upstream connections warm to cut time to first token, then routes each request across providers of the same model by price, latency, throughput, and availability, with automatic failover. Every call goes to an official provider endpoint, billed at their prices.

Alibaba Cloud
Tencent Cloud
Volcengine
Baidu AI Cloud
Zhipu AI
Moonshot AI
MiniMax
DeepSeek
Sign up freeRead API docs
Official endpoints · Pay as you go · Zero content logging
MENUOverviewAPI keysModels & pricingOnline playgroundRequest logsRecharge center
☰⌕   Search features or docs◐　♢　FC
Account balance$1,284.60↑ 18.2%
Requests428,590↑ 12.8%
Service healthAll systems operational
99.72%
Usage trend
☰•••
Account balance$1,284.60↑ 18.2%
Requests428,590↑ 12.8%

One API key, routed across China's leading model providers

AlibabaCloudAlibaba CloudTencentCloudTencent CloudVolcengineVolcengineBaiduCloudBaidu AI Cloud
Why production needs more than one provider

Production workloads should not carry single points of failure

Provider fluctuations should not become business interruptions. From availability and model choice to changing prices, production applications need an alternative route they can take at any time, not one provider they always depend on.

One outage can disrupt your product

Even established providers experience outages, rate limits, and slowdowns. Without an alternative route, an upstream issue can make product features unavailable.

Every new model means a new integration

Providers differ in endpoint shape, keys, accounts, and billing semantics. Switching model or provider can mean rebuilding both the integration and the reconciliation.

A fixed route can cost more

Prices for the same model differ across providers and change over time. Staying with one provider can mean missing a better price elsewhere.

The InOneAPI approach

Leading providers. One smart entry point.

InOneAPI is a purpose-built model gateway, not a forwarding address: one protocol and auth surface, live route selection against your priorities, request-level tracing and token metering, and circuit breaking on failure — all billed at official provider prices.

Automatic failover reduces disruption

Routing responds to latency, throughput, and error rates, shifting between available providers for the same model. Repeatedly failing providers are taken out and probed for recovery.

Flagship models, verified one by one

We connect only flagship models from each leading provider, verified for protocol compatibility and billing semantics before they go live — across text, image, video, and audio.

Lowest available price, automatically

Price-first routing selects the lowest-priced available provider for the same model. Choose speed-first or balanced routing when your business priorities differ.

Core capabilities

Built for production model access

From model coverage and failover to cost control, bring reliability, transparency, and content privacy to every request.

Flagship models, one integration

Access flagship models from China's leading providers across text, image, video, and audio through one API. Choose by use case, not by whoever has the longest catalog.

Routing on your terms

Prioritize price, response speed, or a balance of both. Routes adapt as provider conditions change.

Official endpoints, itemized costs

Requests go to official provider endpoints at their prices, with no markup. Tokens are metered locally using each provider's own algorithm, so every line can be reconciled.

Resilience across providers

Use multiple providers for the same model. Dynamic routing and automatic failover reduce the impact of upstream slowdowns and errors.

Team access and budgets

Manage workspaces, projects, and sub-keys with budget limits, model restrictions, and rate limits.

Zero content logging

Zero content logging is standard across the platform. Prompts and responses are streamed and metered in memory, without persisting their content or requiring opt-in.

In-house gateway, faster first token

Price, speed, and balanced routing

Native OpenAI and Anthropic protocols

Request-level trace diagnostics

Circuit breaking with recovery probes

Error-rate and budget alerts

RPM, TPM, and concurrency limits

One-click agent tool setup

24/7 human support

How it works

What the gateway does with each request

Smart routing

Set the goal. Let routing adapt.

Prices and performance change, so routes should too. Choose price-first, speed-first, or balanced routing, and the gateway selects across providers of the same model by price, first-token latency, throughput, and availability.

Price-first finds the lowest-priced route

Select the lowest-priced available provider for the same model at current prices. Routing adjusts as prices change, reducing manual comparisons and switching.

Respond to performance changes

Routing uses latency, throughput, and error rates to shift traffic as performance drops, and takes repeatedly failing providers out until a probe recovers them.

Pricing and transparency

Smart routing without a price premium

Choose flexibly among leading providers while paying their official prices. Tokens are metered locally with each provider's own algorithm, and the selected provider, usage, and cost are recorded per request, so a bill can be checked line by line against the provider's price.

Request-level tracing

Use X-Gateway-Trace-ID to investigate timing, retries, and usage across the request path.

Itemized costs

Verify each request against the selected provider’s price and the usage recorded.

Production support

Production support. Content stays private.

Multi-provider routing reduces single-provider dependence, while workspaces, projects, and sub-keys set access boundaries and budgets. Zero content logging is standard: meter and debug calls without persisting prompts or responses.

Issues surface, then stop

Error-rate and budget anomalies create alert events, and persistently failing providers are removed from routing instead of keeping live traffic.

Zero content logging is standard

No privacy setting to enable. Content is processed only in memory. Request metadata such as usage, cost, and timing supports billing and diagnostics without prompt or response bodies.

Quick integration

Change one baseURL, reach China's top models

Text requests are natively compatible with OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages. Image and audio reuse OpenAI endpoint shapes, and video runs through unified endpoints. Swap the baseURL and API key and keep running.

View integration docs
quickstartCode examples
# -H "X-APP-NAME: Your App Name"
# -H "X-APP-URL: https://your-app.example.com"
curl "https://api.inoneapi.com/v1/chat/completions" \
  -H "Authorization: Bearer $INONEAPI_API_KEY" \
  -H "Content-Type: application/json" \
  --data '{
  "model": "YOUR_MODEL_ID",
  "messages": [
    {
      "role": "user",
      "content": "Hello"
    }
  ]
}'

Give every model call a second path

Sign up, create an API key, and change one baseURL to get started. Pay as you go, with provider official prices you can reconcile line by line and no content stored.

Sign up free
InOneAPI · Smart routing gateway for China's top models, at provider prices