Okta has unveiled a plan to reduce the cost of using artificial intelligence (AI) agents by identifying and analyzing the "tool tax" associated with each model call made. The company claims its Identity-Scoped Model Context Protocol tool lists can significantly decrease these costs, which are generated as models consider tools exposed by a server.
The tool tax refers to the amount of tokens consumed during an AI agent's operation, including schemas, names, descriptions, and parameters for various tools that come into play. By analyzing this data, Okta is able to identify areas where model calls can be optimized without significantly impacting performance. This is particularly important in industries such as cybersecurity, where efficiency gains are crucial.
According to Okta, the MCP scoping tool lists can halve AI agent token costs by reducing unnecessary computations and improving overall model performance. The company hopes that this solution will benefit businesses, startups, and organizations across various sectors that rely on AI-powered applications.