AI Gateway
One place to send every AI request from your product.
Learn moreOne gateway for OpenAI, Anthropic, Gemini, Groq, and more. Automatic failover, model routing, and spend controls are built in.
Built for indie hackers, startups, and AI SaaS teams.
Value before pricing. These are the outcomes teams get when they stop wiring providers one by one.
One place to send every AI request from your product.
Learn moreKeep features online when a provider slows down or fails.
Learn moreInspect request logs so you can debug failures and understand usage.
Learn moreSet budgets so one workflow cannot surprise you on the bill.
Learn moreConfigure models, backups, and outputs without redeploying.
Learn moreShow answers as they generate for a faster product feel.
Learn moreOpenAI-compatible chat completions plus native ModelRiver endpoints.
Learn moreForward production events to local development without exposing your machine.
Learn moreCapabilities first. Usage limits second. Upgrade when reliability and volume matter more.
Launch an AI gateway and prove the workflow before you spend.
CapabilitiesNo credit card required.
Run production AI with failover, cost control, and room for real traffic.
CapabilitiesChoose Core when AI reliability matters.
Volume pricing and dedicated help for teams shipping AI at scale.
CapabilitiesProduction status: 99.9% uptime
Trust signals from how ModelRiver actually works — not vanity metrics.
Connect your own keys. Swap models in the dashboard without redeploying.
What you get on each plan. Limits change; the gateway stays the same.
| Feature | Free | Core | Enterprise |
|---|---|---|---|
| AI Gateway | Yes | Yes | Yes |
| Unified API | Yes | Yes | Yes |
| Automatic Failover | Yes | Yes | Yes |
| Observability | Yes | Yes | Yes |
| Workflows | Yes | Yes | Yes |
| Streaming | Yes | Yes | Yes |
| Spend Limits | Yes | Yes | Yes |
| Teams | 1 seat | Unlimited | Unlimited |
| CLI | Yes | Yes | Yes |
| Support | Community | Priority email | Dedicated |
| Monthly requests | 25,000 | 250,000 | Custom |
Switch in minutes
Change one endpoint and start using ModelRiver. Keep your SDK, bring your own keys, and add failover when you are ready.
Keep your existing SDK
OpenAI-compatible chat completions — no rewrite required.
Point to ModelRiver
Migration often takes minutes, not weeks.
Add backups when ready
No one-vendor lock-in — switch models and providers anytime.
What you get
You still pay providers directly. ModelRiver is the control layer that keeps your product reliable and flexible.
Avoid one-vendor lock-in
Automatic provider failover
Cost control with spend limits
Request logs and centralized monitoring
One API for every provider
Straight answers for teams deciding how to build AI into their product.
Direct OpenAI integrations are fine for experiments, but they lock your product to one provider and push failover, monitoring, and cost control onto your team. ModelRiver gives you one gateway for many providers so you can switch models later and stay online when a provider has issues.
Yes. You connect your own provider keys in the console. ModelRiver does not resell tokens. You pay providers directly and pay ModelRiver for the platform layer.
If a provider is slow, unavailable, or rate limited, ModelRiver can retry through another provider you configured. That protects customer experience without custom fallback code for every model.
Yes. That is a core reason to use ModelRiver. Move traffic between OpenAI, Anthropic, Gemini, and others from the dashboard without rebuilding your product around a new API.
Set project and per-model budgets. ModelRiver can warn as you approach the limit and block over-budget live requests before they hit a provider bill.
Requests route through ModelRiver to the providers you choose. Use your own keys, and review our docs and policies for details on how traffic and logs are handled.
You're not locked into a single model provider. Your app talks to a stable ModelRiver API, and you keep provider accounts under your control. Leave anytime and reconnect providers directly if you want.
Many teams start by pointing an OpenAI-compatible client at ModelRiver and swapping the base URL. Full workflow setup usually takes minutes to a couple of hours, not a multi-week rewrite.
A request is an AI call routed through ModelRiver, including streaming calls and retries handled as part of your setup. Each plan shows included monthly requests and published overage rates where applicable.
Yes. Upgrades take effect immediately. Downgrades take effect at the next billing cycle. Mid-cycle upgrades are prorated.
Subscriptions are billed monthly and are non-refundable except as stated in our refund policy. Start on Free to validate fit before upgrading.
Connect your providers in minutes. Automatic backups, spend limits, and one place to manage every model — no credit card required.