Ramp introduces Router to manage AI model access

Corporate expense management firm Ramp has launched a new artificial intelligence service designed to facilitate the routing of queries between various large language models. The platform, named Router, allows businesses and individual users to access and switch between different AI providers through a single application programming interface. The company stated that it has been utilising the internal routing infrastructure for its own operational needs for the past three years before making it available to the wider market.

The service is currently restricted to users within the United States. Ramp has announced that the platform will be free of charge for the remainder of 2026, although users will still be responsible for the underlying costs of AI model inference. To encourage early adoption, the company is offering a launch credit of 26 dollars. No specific pricing structure has been disclosed for the service beyond the current year. While the offering functions similarly to the existing OpenRouter platform, Ramp’s current selection of available models is more limited than that of its competitor.

Router provides access to models developed by OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai. The platform includes several strategic tools to assist customers in directing AI requests according to specific preferences. These features allow users to prioritise flexible usage tiers from model providers or enable the system to select models based on up to three user-defined benchmarks. Additionally, the service permits users to route only complex problems to more expensive models or to test different models easily without the need for manual switching. A dedicated dashboard is provided to monitor token spend, costs, latency, and fallback attempts.

Data handling policies for the service include an opt-out retention mechanism. By default, the platform records model inputs, outputs, and tool calls for a period of one year. Ramp asserts that personally identifiable information is removed before this content is used to improve the product. The move into model routing presents a dual opportunity for the company. It allows Ramp to participate in the expanding AI inference market while offering its existing client base a service that integrates with its current products, which include AI token usage monitoring and spend management.

If the platform gains significant traction as a testing environment for AI models, Ramp may establish enduring relationships with AI laboratories and inference providers globally. This could assist the company in acquiring new customers and creating an additional entry point for its expense management products. Ramp recently raised 750 million dollars in a funding round that valued the company at 44 billion dollars in June.