Ramp is not a company most developers think of when shopping for AI infrastructure tools. That might be about to change. The corporate expense management platform has launched Router, its own AI model routing service that lets developers and companies switch between multiple large language models through a single API. The move puts Ramp in direct competition with OpenRouter, and follows Stripe’s own push into the AI inference space.
Router is available now, but only in the United States. It’s free to use for the rest of 2026, though users still pay their own model inference costs. There’s a $26 credit included at launch. Ramp has not said what it will cost in 2027, which is something anyone building on top of it should flag before going too deep.
The model selection is narrower than OpenRouter’s current catalog, but covers the major names. Router gives access to models from OpenAI, Anthropic, DeepSeek, xAI, Nvidia, Moonshot, Minimax, and Z.ai. What makes it more interesting than a basic proxy layer is the routing logic on top. Ramp offers several strategies that let users control how requests get directed:
- Route queries based on up to three user-specified benchmarks
- Prefer providers’ flex usage tiers to reduce cost
- Send only complex or difficult requests to more expensive models
- Test different models without switching API configurations
There’s also a dashboard showing token spend, latency, cost, fallback attempts, and other usage details. For teams already managing AI budgets, that visibility is genuinely useful, and it’s where Ramp’s existing product experience in expense and spend management starts to show.
One thing worth watching is the data policy. Router records model inputs, outputs, and tool calls for one year by default. Users have to actively opt out. Ramp says it strips personally identifiable information before using that data to improve the product, but an opt-out default on data retention is the kind of thing that will bother enterprise security teams. It’s not disqualifying, but it needs to be on your checklist before you sign up.
Ramp says it has been running this routing infrastructure internally for three years. That’s a credible backstory. It also suggests Router isn’t a speculative side project but something the company actually depends on. Still, going from internal tooling to a public developer product is a real transition, and OpenRouter has a significant head start in model breadth and developer mindshare.
So why does this matter beyond another entrant in the routing space? Ramp raised $750 million at a $44 billion valuation in June. It has a large existing customer base of finance and operations teams who already use its spend management products. Router gives those customers a natural on-ramp to AI tooling that connects directly to what they already track, AI token costs. And for Ramp, it creates a relationship with AI labs and inference providers that could translate into new enterprise sales channels over time.
This is not a product that will displace OpenRouter for developers who are already embedded in that ecosystem. But for companies already on Ramp who want routing and spend visibility in one place, the value proposition is straightforward. Watch whether the pricing announced early next year is competitive enough to make that convenience worth paying for.
