OpenAI has revealed its support for the MCP protocol, an open-source technology developed by Anthropic PBC, a key competitor of ChatGPT. The MCP protocol is designed to enable large language models (LLMs) to interact with external systems and access data within those environments.
Sam Altman, CEO of OpenAI, announced this development on Wednesday. It’s worth noting that the MCP protocol was originally created by Anthropic PBC, which is one of the most competitive startups in the field against ChatGPT's developer.
Currently, MCP support has been integrated into OpenAI’s Agents SDK, an open-source toolkit designed to help developers build AI agents. Over the coming months, OpenAI intends to incorporate MCP into ChatGPT’s desktop client and Responses API, the latter being a programming interface that allows developers to access its language models.
By linking with external systems, companies can significantly enhance the practical applications of their language models. For example, retailers can connect a language model to a product database to generate shopping recommendations. Previously, creating such integrations required substantial effort, but Anthropic’s MCP protocol streamlines this process.
The MCP protocol provides software building blocks that allow developers to quickly connect language models to external systems. According to Anthropic, integrating using this protocol can take less than an hour in some cases.
In addition to retrieving data from external systems, MCP also enables language models to perform operations within those systems. For instance, a language model optimized for coding tasks could use the MCP protocol to execute configuration scripts on cloud instances. Similarly, an AI-driven marketing tool could feed ad performance metrics into analytics applications. With OpenAI’s adoption of the MCP protocol, users of ChatGPT’s desktop app, Responses API, and Agents SDK will gain access to these capabilities.
This week’s update coincides with the release of the latest version of the MCP protocol. The new version introduces several enhancements, including batch processing capabilities using JSON-RPC technology for more efficient data transfer. It also simplifies the notification process for language models accessing MCP-enabled systems and upgrades the authorization mechanism to OAuth 2.1 for improved security.
Meanwhile, Microsoft, an investor in OpenAI, has launched its own MCP integration tool called Playwright MCP. This tool combines Anthropic’s MCP protocol with Microsoft’s Playwright software, leveraging its web browsing functionality to enable language models to interact with web pages.
Developers can utilize Playwright MCP to automate tasks such as filling out forms. Microsoft also notes that this tool allows coding-focused language models to automate website testing processes.
