Okta is exploring the use of zero-knowledge proof technology for managing AI agents. The company claims that its Identity-Scoped Model Context Protocol (MCP) tool can help reduce the cost of AI agent tokens by minimizing the amount of data required to expose tools used by these agents.
Each model call made by an AI agent involves a list of schemas, names, descriptions, and parameters for every tool exposed by a MCP server. This results in significant overhead, which Okta refers to as the "tool tax". The company says that by applying zero-knowledge proof techniques, it can reduce this overhead, thereby lowering the cost of AI agent tokens.
By employing these technologies, Okta aims to provide its clients with more efficient and cost-effective solutions for managing their AI agents. This move is seen as an important step towards mitigating some of the significant costs associated with developing and deploying AI-powered applications.