Unlocking Savings in Machine Learning: How Okta's MCP Scoping Can Save You Millions
Okta claims that its Model Context Protocol tool, or MCP for short, can help reduce the cost of using artificial intelligence by exposing every aspect of a machine learning model, including schemas, names, descriptions, and parameters. This means that AI agents must consider not only their input data but also the tools exposed by the server where they are running. The resulting prompt overhead known as the "tool tax" is something that Okta wants to minimize.
By using MCP scoping, companies can avoid paying for each tool call separately when making model calls. According to Okta, this means a reduction in AI agent token costs, which can be substantial. In some cases, it could even lead to significant savings. For example, if an AI agent is called with 100 different tools, the total cost of using MCP scoping would instead involve paying for 1,000 individual tool calls.
Okta's approach to MCP scoping is intended to improve efficiency and reduce waste in machine learning workflows. By exposing every aspect of a model, Okta believes that companies can optimize their use of AI resources and get more value out of their investment. The company has already seen positive results from implementing its MCP scoping tool, with some customers reporting reductions in costs by up to 20%.