<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Mcp on Shivasurya</title><link>http://shivasurya.me/tags/mcp/</link><description>Recent content in Mcp on Shivasurya</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sat, 19 Jul 2025 00:00:00 +0000</lastBuildDate><atom:link href="http://shivasurya.me/tags/mcp/feed.xml" rel="self" type="application/rss+xml"/><item><title>Rethinking MCP or Tool Calling Through Permission Based System</title><link>http://shivasurya.me/2025/07/19/mcp-permission-system/</link><pubDate>Sat, 19 Jul 2025 00:00:00 +0000</pubDate><guid>http://shivasurya.me/2025/07/19/mcp-permission-system/</guid><description>&lt;p>Model Context Protocol (MCP) and Tool Calling are revolutionizing the application layer of Large Language Models (LLMs), enabling AI to autonomously operate tools and MCP servers to complete tasks. While these capabilities are typically distributed as npm packages or hosted remotely, this distribution method poses potential security risks through malicious code.&lt;/p>
&lt;p>Despite these concerns, Tool Calling and MCP add significant value to AI applications. For instance, Windsurf IDE demonstrates excellent integration by leveraging various tools for file operations, diff viewing, and command execution. Users can configure their own MCP packages and servers, allowing models to control these tools effectively.&lt;/p></description></item></channel></rss>