Join daily and weekly newsletters to obtain the latest updates and exclusive content to cover the leading artificial intelligence in the industry. Learn more
Model context protocol (MCP) – an open standard designed to help artificial intelligence factors interact smoothly with tools, data and interfaces – reaches an important milestone. Today, the developers behind the initiative ended an updated version of MCP specifications, as they presented the keys to make artificial intelligence agents safer, capable and interim operational.
(Update: In a very big step, Openai, the industry leader in AI Tolide, followed MCP today by saying it also adds support to MCP through its products. CEO Sam Altman He said that support is available today At Openai’s Agents SDK and support the desktog application in ChatgPT and API responses will come soon.)
In a noticeable step, two days ago Microsoft He also recently announced the support of MCP alongside this version, including a new launch Playwright-MCP server This allows artificial intelligence factors such as Claud to browsing the web and interacting with sites using the access to Chrome.
“This new version is a big leap forward for the Tail tools contacts,” said Alex Albert, the main contributor to the MCP project. In a publication On Twitter. “The presence of Microsoft is building infrastructure in the real world, moreover, it indicates the rapid development of this ecosystem.”
What is new in the updated mcp Issue?
March 26 update brings many important changes at the protocol level:
- OATH 2.1 delegation frameworkIt adds a strong standard to secure a dealer server connection, especially in HTTP transfers.
- Transfer HTTP to the restThe oldest HTTP+SSE setting is replaced, allowing data flow in actual time, with better compatibility.
- Json-RPC renewalCustomers are allowed to send multiple requests in one case, improve efficiency and reduce access time in agent tool reactions.
- Tool commentsIt adds rich descriptive data to describe the tool behavior, which allows more discovery and thinking by artificial intelligence agents.
Figure 1: CLADE desktop using the MCP theatrical writer for transmission and description of Datasette.io, which indicates web automation backed by the form of the context of the context of the model.
The protocol uses a standard base JSON-RPC 2.0, with a structure with layers separating basic transportation, life cycle management, server features (such as resources and demands) and customer features (such as samples or registration). Developers can choose and choose the components that must be implemented, depending on their use.
Microsoft’s contribution: McP browser automation
As we mentioned, two days ago, Microsoft released playwright-MCPServer wraps a powerful browser automation tool in the MCP standard. This means that artificial intelligence agents like Claude can now do more than conversation – they can click, books, browsing and interacting with the web like a real user.
Built on the access to Chrome, CLADE allows access to the content of the page and described them in a human readable form. It includes a set of tools available:
- navigation: Browser_Navity, Go_Back, Go_Forward
- entrance: Browser_Type, Browser_CLICK, Browser_Press_KEY
- Shots: Browser_snapshot, browser_screenshot
- Interactions based on elements Using accessible accessories
This transforms any AI agent that is compatible into a test automation robot or QA or Navigator data.
People love MCP and we are excited to add support through our products.
Available today in SDK agents and support for the next Chatgpt Desktop App + API!
Sam Al -Taman (Sama) March 26, 2025
Easy setting: Users simply add the theatrical writer as a matter of Claud_Desktop_config.json, and the Claud Desktop app will learn about the tools at the time of operation.
The largest image: widespread intercourse
Figure 2: The standard design of MCP enables developers to implement the layers they need only, while maintaining compatibility.
man It was first presented MCP In late 2023 to solve the growing pain point: Artificial intelligence factors need to interact with the tools of the real world, but each application speaks different “language”. MCP aims to fix this by providing a common protocol to describe and use tools through ecosystems.
With support from Antarbur, Langchain, and now Microsoft, MCP appears as a dangerous competitor to become the standard layer to reach the connection. Since MCP was first launched by Anthropor, the questions have remained if the largest competitor to anthropologist, Openai, has been supporting the protocol. Of course, Microsoft, a great ally of Openai, was another question mark. The fact that both players have supported the protocol showing momentum is construction between institutional societies and open source. Openai itself has opened its ecological system on agents, including with the latest agents announced a week ago – and This step has strengthened support for Openai API formatGiven that others like anthropoor and Google have fallen into the queue. So with Openai and MCP API format, the support has seen a great victory over the past two weeks.
“We enter the era of artificial intelligence protocol,” Tweet Alexander Doria, co -founder of AI Startup Pleias. “Thus, the agents will do in reality Do Things. “
What next?
With the MCP 0.2 version and the concrete microsoft support, the foundation is laid for a new generation of agents who can think and act safely and flexible across the stack.
Figure 3: OATH 2.1 Delegate flow in the context of the context of the form (MCP)
The big question now is: Do others follow? If Meta, Amazon or Apple Sign, MCP may soon become a global “language” for artificial intelligence procedures.
Currently, it is a big day for the agent’s ecosystem – which brings the promise of interfering operating from artificial intelligence closer to reality.
https://venturebeat.com/wp-content/uploads/2025/03/nuneybits_Abstract_art_of_robots_chatting_on_instant_messenger__44be2d4d-a3e3-408b-b11f-a0bf322c2a61.webp?w=1024?w=1200&strip=all
Source link