Okta has released an update to its identity-scoped Model Context Protocol (MCP) tool that addresses one of the biggest pain points in artificial intelligence - pricing. The company's tool, designed to help developers build more efficient and scalable AI models, now includes a feature called "tool tax" which aims to reduce costs associated with tokenization.
The tool allows developers to include schemas, names, descriptions, and parameters for every tool exposed by an MCP server during model training. This results in significant prompt overhead as each model call made by the AI agent can include these details. Okta refers to this as "tool tax", which represents the tokens consumed by a model when it considers tools.
By implementing this feature, developers using Okta's MCP tool will be able to reduce their AI agent token costs and improve overall efficiency in developing and deploying AI models. The tool is designed to work seamlessly with existing training data and can handle large-scale deployments. Okta plans to continue refining its tool to further minimize the "tool tax" and make it more user-friendly, making it easier for developers to build high-performance AI models without the added cost of unnecessary overhead.