Ramp, the corporate expense management platform, has launched Router, its AI model routing service. It lets users and companies send requests through an API and switch between large language models, and Ramp says it has been using the router internally for its own AI needs for the past three years. The service is only available in the United States. It is free to use for the remainder of 2026, though users still pay model inference costs, and it comes with a $26 launch credit. The company did not disclose pricing beyond 2026.

Router works similarly to OpenRouter, which offers many more model options. Ramp‘s current Router provides access to models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai. Ramp also offers several routing strategies: for example, one lets users set a preference for providers’ flex usage tiers, another lets Router choose a model based on up to three user-specified benchmarks, and another routes only difficult problems to more expensive models. Users can also test models easily without switching configurations. The attached dashboard shows token spend, cost, latency, fallback attempts, and other diagnostics.

On data, Router has an opt-out retention policy. Model inputs, outputs, and tool calls are recorded by default for one year, but the company says it removes personally identifiable information before using stored content to improve the product. For Ramp, the router is a two-pronged opportunity: tapping the growing AI inference market and giving existing clients a model routing tool that fits alongside its token usage monitoring and token spend management. If Router becomes as popular a testing arena as OpenRouter, Ramp could also build long-term relationships with AI labs and inference providers, potentially attracting new customers and bringing expense management products through a new entry point. Ramp raised $750 million at a $44 billion valuation in June.