OpenAI announced that its core Responses API now officially supports the Model Context Protocol (MCP), injecting new vitality into the development of AI agents. This major update significantly lowers the threshold for developers to connect AI models with external tools, marking a crucial step forward for OpenAI in promoting openness and collaboration within the AI ecosystem.
The MCP protocol, proposed as an open-source standard by Anthropic in 2023, aims to unify the interaction methods between AI models and external tools/services, akin to HTTP in the web domain. Through MCP, developers can easily connect AI models to e-commerce platforms, communication services, or other third-party systems, enabling more complex workflows. OpenAI's update allows its Responses API to connect any MCP server with just a few lines of code, greatly enhancing development efficiency.
OpenAI also announced its inclusion in the MCP Advisory Committee, indicating its commitment to driving standardization and open collaboration in the AI industry. This move not only strengthens OpenAI's competitiveness in the field of AI agents but also provides developers with more flexible tool integration options.
Responses API Upgrade: Comprehensive Function Enhancement
OpenAI's Responses API is specifically designed for developing intelligent, action-oriented AI agents. This update not only introduces MCP support but also adds image generation, code interpreter, and optimized file search functions, further enriching the capabilities of the API. Developers can now achieve the following scenarios via MCP:
E-commerce Automation: Just nine lines of code are needed to connect the agent to Shopify, facilitating order processing and inventory management.
Real-time Communication: With 13 lines of code, the agent can interact with cloud communication platforms like Twilio, supporting SMS or voice notifications.
Complex Task Handling: The optimized file search function enables agents to efficiently handle multi-step tasks.
These features went live on May 21, 2025. Specific implementation methods and pricing can be referenced in the official OpenAI documentation.
Industry Trend: MCP Becomes the New Standard for AI Agents
OpenAI’s support for MCP aligns closely with industry trends. Recently, tech giants like Google and Microsoft have also announced their support for the MCP protocol, pushing it to become the de facto standard for AI agent development. For example, Google has integrated MCP into its Gemini model and related SDKs, while Microsoft offers MCP support in Azure AI Foundry and Co-pilot Studio. This cross-platform collaboration indicates that MCP is breaking down barriers between AI models and tools, allowing developers to freely choose combinations of models and tools.
Social media discussions further reflect the potential of MCP. Developers generally believe that the standardization characteristics of MCP will significantly reduce development costs, especially in scenarios requiring interaction with complex external systems. Some comments note: “With OpenAI, Google, and Claude all supporting MCP, the future of AI agent development looks bright!” However, some voices mentioned that Anthropic’s Claude leans more toward application layers in MCP support, with slightly less depth in API-level integration.
On social media platforms, developers reacted enthusiastically to OpenAI’s MCP support update. Many stated that MCP support made agent development simpler and more efficient, particularly the feature allowing connection to external tools with just a few lines of code, which greatly improved the development experience. One developer commented: “After OpenAI’s API supported MCP, developing agents feels as simple as building blocks!” Additionally, MCP’s cross-platform compatibility allows developers to switch models more flexibly without having to rebuild existing systems.
However, some developers pointed out that although OpenAI and Google provide deep MCP support at the API level, the industry still needs more tools and services to adapt to MCP to fully realize its potential.