Antroppac has updated the Multi-Modal Communication Protocol (MCP), which reinforces the interconnection and security between artificial intelligence (AI) agents. Specifically, after the open AI, Microsoft (MS) integrates MCP into the AI agent, and the protocol is becoming a brand new standard for AI agent linkage.
Antropic announced on the twenty sixth (local time) that it has launched an MCP update version that more efficiently and safely enhanced the interaction between AI agents and tools through X (Twitter).
This update has been added with the ‘OAUTH 2.1’ -based certification framework, which has a robust standard that may protect the communication between agents and servers more safely in HTTP -based transmission.
As well as, the stream mube HTTP transmission function replaces the present HTTP+SSE settings, enabling real -time bidirectional data flow and improving system compatibility.
As well as, the JSON-RPC betrayal function has been added, allowing the client to send several requests without delay, increasing the interaction efficiency between the agent and the tool, and the delay time.
Lastly, the ‘Annotations’ function was introduced so as to add a wealthy metadata to elucidate the motion of the tool, which allowed the AI agent to explore and use the tool creatively.
The protocol relies on modular JSON-RPC 2.0 and adopts hierarchical architecture that separates client functions resembling core transmission, life cycle management, resources and promptes resembling server functions resembling sampling and logging. This enables developers to pick out and implement only the required components in response to their use.
At the identical time, Microsoft announced that it could introduce MCP to its own AI agent construction.
Microsoft has launched the ‘Playwright’, a browser automation tool, ‘Playwright’ to launch ‘Play Lighter-MCP Server’ to assist LLM interacts with web pages.
The server has built the playwriter in response to the MCP standard, and now supports AI agents, resembling Claude, in order that they’ll click, input, web search and interaction beyond easy conversations.
This integration relies on the Chrome accessible tree, allowing Claude to research and explain the content of the net page in a form that an individual can read.
Open AI also announced that it could introduce Antropic’s MCP to construct AI agent the day before. Open AI said that MCP will likely be available in its open agent SDK, and will likely be provided by the chat GPT desktop app and the Responses API.
Antropic introduced MCP in November last yr to resolve the issues that arise when the AI agent interacts with the actual tool.
To unravel the issue that every application uses different languages, the MCP provides protocols that might be described and utilized in a standardized manner in various ecosystems. Developers can provide data through the MCP server and develop MCP clients resembling AI Agents and connect them to the server when needed.
Since Antropic released MCP as an open source, many corporations resembling Langchain, Block, Apollo, Ripplet, Codium, and Source Graph have been popular, adding MCP to their very own platforms. It’s now supporting MCPs to open AI and Microsoft (MS), which is more more likely to grow to be an industry standard for interconnection between AI agents.
“This latest MCP version means a giant progress in communication between agents and tools,” said Alex Albert Antrofac, a researcher on the MCP project.
By Park Chan, reporter cpark@aitimes.com
