Okta has unveiled its plan to dominate the AI agent token market by introducing a new identity-scoped Model Context Protocol tool. According to the company, this protocol can reduce AI agent token costs significantly by listing tools in an identity scope that is scoped to an individual user's model. This means that each model call made by an AI agent can include schemas, names, descriptions, and parameters for every tool exposed by a MCP server.
The "tool tax" as Okta calls the resulting prompt overhead refers to tokens consumed as a model considers tools. Each token consumption is tied to specific models, making it costly for users to list numerous tools in their identity scope. However, with Okta's new protocol, this cost can be significantly reduced by listing only those tools that are actually used.
By reducing AI agent token costs, Okta aims to make its platform more appealing to businesses and individuals alike. The company claims that its approach will increase user adoption of its MCP tool and provide a more streamlined experience for users. While the specifics of how Okta's protocol works and whether it will gain traction in the market remain unclear, the introduction of this new feature is seen as a significant step forward by the company.