Join our daily and weekly newsletter for the latest updates and exclusive content on the top AI coverage. Learn more
The Model Context Protocol (MCP) – a rising open standard designed to help AI agents interact seamlessly with tools, data, and interfaces – only hit a significant milestone. Today, the developers behind the initiative have won an updated version of the MCP spec, which introduces major upgrades to make AI agents safer, capable of being able to.
[Update: In a very significant move, OpenAI, the industry leader in generative AI, followed the MCP announcement today by saying it also is adding support for MCP across its products. CEO Sam Altman said the support is available today in OpenAI’s Agents SDK and that support for ChatGPT’s desktop app and the Responses API would be coming soon.]
In a well -known move, two days ago Microsoft The support for the MCP was also recently announced next to this release, along with the launch of the new Playwright-MCP server It allows AI agents such as Claude to browse the web and interact with sites using Chrome accessing tree.
“This new version is a major leap forward for agent-tool communication,” Alex Albert, a major contributor to the MCP project, said In a post on twitter. “And having a Microsoft Building real-world infrastructure at the top shows how fast this ecosystem is growing.”
What is new to that -updated MCP Version?
March 26 update brings many important changes to protocol levels:
- Oauth 2.1 Based on Authorization based on Oauth: Adds a stable standard for securing agent-server communication, especially in HTTP-based transportation.
- Streamable http transport: Replacing older http+sse setup, enabling real-time, bidirectional data flow with better compatibility.
- JSON-RPC batching: Allows clients to send multiple requests to a go, improving efficiency and reducing latency in agent contacts.
- Tool annotations: Adds a rich metadata for describing the tool behavior, enabling the more imaginative discovery and reasoning of AI agents.
Figure 1: Claude Desktop using the Playwright-MCP to navigate and describe Datasette.io, which shows web automation enabled by the model context protocol.
Protocol uses a modular base of JSON-RPC 2.0, with a layered architecture that separates basic transportation, lifecycle management, server features (such as resources and signals) and client features (such as sampling or logging). Developers can choose and choose which ingredients should be implemented, depending on their use case.
Microsoft Contribution: Browser Automation by MCP
As mentioned, two days ago Microsoft released In the playmateMcPa server that wraps a strong browser automation tool with MCP standard. This means that AI agents like Claude can now do more than just talk -they can click, type, browse and socialize on the web like a real user.
Built on Chrome access tree, integration allows Claude to access and describe the contents of the page in a human readable form. The available toolet includes:
- Navigation: browser_navigate, go_back, go_forward
- Input: browser_type, browser_click, browser_press_key
- Snapshots: browser_snapshot, browser_screenshot
- Elements-based contacts Using accessing descriptors
It turns out any following AI agent in a test automation bot, QA assistant or data navigator.
People like the MCP and we're excited to add support to our products.
Now available to SDK agents and support for ChatGPT Desktop App + API Responses coming up!
– Sam Altman (@sama) March 26, 2025
Setup is easy: Users simply add a playwright as a command to Claude_Desktop_config.json, and the Claude Desktop app will recognize runtime tools.
The bigger picture: Interoperability in the scale
Figure 2: The modular MCP design gives developers to implement only the layers they need, while maintaining compatibility.
Anthropic first introduced McP In the late 2023 to resolve a growing point of illness: AI agents need to contact real-world tools, but each app speaks a different “language.” The MCP aims to adjust that by providing a common protocol for describing and using tools throughout the ecosystem.
Backing from the anthropic, langchain, and now Microsoft, the MCP is emerging as a serious contender for being the usual layer of agent coherence. Since the MCP was launched first by anthropic, the questions lasted if Anthropic's largest competitor Openai, would support the protocol. And of course, Microsoft, a huge allies of Openai, is another question mark. The fact that both players support the protocol shows momentum is the development of business communities and open resources. The Openai itself opens the ecosystem around the agents, with the latest agents announced by the SDK a week ago – and Moving has stable support around Openai APII formats that become a standardBecause that others like anthropic and google fell on the line. So in Openai's API formats and the MCP has both seen support, the standardization has seen a huge win over the past few weeks.
“We entered during the AI protocol,” tweets Alexander Doria, the co-founder of the AI startup pleias. “This is how really agents are do things. “
What's next?
With the release of MCP 0.2 and Microsoft's tangible support, the grounding is placed for a new generation of agents who can think and act safely and flexible throughout the clamp.
Figure 3: Oauth 2.1 Authorization Flow to Model Context Protocol (MCP)
The big question now is: Will others follow? If the Meta, Amazon, or Apple signing, the MCP can be the universal “language” of AI actions.
So far, it's a big day for the agent's ecosystem – the one that brings the promise of AI's interoperability closer to reality.
