<?xml version="1.0" encoding="UTF-8"?>
<!-- Generated by scripts/build-server-pages.ts from data/news.json. Do not edit. -->
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Axis MCP news</title>
    <link>https://axismcp.dev/news/</link>
    <atom:link href="https://axismcp.dev/news/feed.xml" rel="self" type="application/rss+xml"/>
    <description>Model Context Protocol news with a short note from Axis MCP, plus the catalog changelog.</description>
    <language>en</language>
    <lastBuildDate>Mon, 12 Oct 2026 12:00:00 GMT</lastBuildDate>
    <item>
      <title>Setup guides for Claude Desktop, Claude Code and Codex, Cursor and VS Code</title>
      <link>https://axismcp.dev/guides/</link>
      <guid isPermaLink="false">axismcp-news-2026-10-12-setup-guides-for-claude-desktop-claude-code-and-codex-cursor-and-vs-code</guid>
      <pubDate>Mon, 12 Oct 2026 12:00:00 GMT</pubDate>
      <category>Our changelog</category>
      <description>Axis MCP catalog. Each guide shows where the client reads its config, how it handles keys, and how to check a server loaded. They draw on the same per-client examples as the server pages.</description>
    </item>
    <item>
      <title>26 current servers added to the catalog</title>
      <link>https://axismcp.dev/servers/</link>
      <guid isPermaLink="false">axismcp-news-2026-10-12-26-current-servers-added-to-the-catalog</guid>
      <pubDate>Mon, 12 Oct 2026 12:00:00 GMT</pubDate>
      <category>Our changelog</category>
      <description>Axis MCP catalog. Vendor-official, actively maintained servers, each checked against its upstream source on Oct 11 with setup examples for six clients and a reviewed write-up. Retired listings now point to their maintained successors.</description>
    </item>
    <item>
      <title>A page for each server</title>
      <link>https://axismcp.dev/servers/</link>
      <guid isPermaLink="false">axismcp-news-2026-10-11-a-page-for-each-server</guid>
      <pubDate>Sun, 11 Oct 2026 12:00:00 GMT</pubDate>
      <category>Our changelog</category>
      <description>Axis MCP catalog. Every configurable server now has its own page with a reviewed write-up, example tools, setup for each client and the sources it was checked against. Archived and deprecated listings stay reachable for existing users but are not offered to search engines.</description>
    </item>
    <item>
      <title>About, Privacy and Contact pages</title>
      <link>https://axismcp.dev/about</link>
      <guid isPermaLink="false">axismcp-news-2026-10-11-about-privacy-and-contact-pages</guid>
      <pubDate>Sun, 11 Oct 2026 12:00:00 GMT</pubDate>
      <category>Our changelog</category>
      <description>Axis MCP catalog. The site now says who runs it, what it collects and how to send a correction or a server for review.</description>
    </item>
    <item>
      <title>MCP was always a bad idea?</title>
      <link>https://simonwillison.net/2026/Sep/20/hn-49779718/</link>
      <guid isPermaLink="false">axismcp-news-2026-09-20-mcp-was-always-a-bad-idea</guid>
      <pubDate>Sun, 20 Sep 2026 12:00:00 GMT</pubDate>
      <category>Explainer</category>
      <description>Simon Willison. Our take: Willison's reply: if you don't want an agent with unrestricted shell and internet access, MCP gives you control over which services it reaches, auth that keeps API keys away from the agent, and audit logging. That is the practical case for installing servers instead of handing an agent raw keys. Our auth and security guide covers where each client keeps those credentials.</description>
    </item>
    <item>
      <title>The New MCP Roadmap</title>
      <link>https://blog.modelcontextprotocol.io/posts/mcp-roadmap/</link>
      <guid isPermaLink="false">axismcp-news-2026-08-22-the-new-mcp-roadmap</guid>
      <pubDate>Sat, 22 Aug 2026 12:00:00 GMT</pubDate>
      <category>Spec and SDKs</category>
      <description>MCP blog. Our take: Next priorities include running local servers over Streamable HTTP and agent identity built on existing standards &quot;rather than pasted API keys and long-lived tokens.&quot; Nothing changes in your config today. Our remote vs local guide explains how the two transports differ now.</description>
    </item>
    <item>
      <title>How Cloudflare detects MCP traffic and helps secure it</title>
      <link>https://blog.cloudflare.com/mcp-security-updates/</link>
      <guid isPermaLink="false">axismcp-news-2026-08-14-how-cloudflare-detects-mcp-traffic-and-helps-secure-it</guid>
      <pubDate>Fri, 14 Aug 2026 12:00:00 GMT</pubDate>
      <category>Advisory</category>
      <description>Cloudflare blog. Our take: Cloudflare Gateway can now flag inspected MCP traffic by its MCP-Protocol-Version header and block direct connections that skip an approved MCP Portal. It cannot see local stdio servers, which never touch the network. If your company runs Gateway, a remote server you add yourself may be blocked unless IT puts it behind the Portal; see remote vs local.</description>
    </item>
    <item>
      <title>Stateless MCP has recaptured my interest (and inspired mcp-explorer and datasette-mcp)</title>
      <link>https://simonwillison.net/2026/Jul/31/stateless-mcp/</link>
      <guid isPermaLink="false">axismcp-news-2026-07-31-stateless-mcp-has-recaptured-my-interest-and-inspired-mcp-explorer-and-datasette-mcp</guid>
      <pubDate>Fri, 31 Jul 2026 12:00:00 GMT</pubDate>
      <category>Explainer</category>
      <description>Simon Willison. Our take: The 2026-07-28 spec drops the initialize handshake and session IDs, so a tool call is one self-describing HTTP request. Willison's mcp-explorer CLI lists and calls a server's tools, a quick way to see what a remote server exposes before you add it. He also argues MCP tools are easier to audit than an agent with a shell.</description>
    </item>
    <item>
      <title>The 2026-07-28 Specification</title>
      <link>https://blog.modelcontextprotocol.io/posts/2026-07-28/</link>
      <guid isPermaLink="false">axismcp-news-2026-07-28-the-2026-07-28-specification</guid>
      <pubDate>Tue, 28 Jul 2026 12:00:00 GMT</pubDate>
      <category>Spec and SDKs</category>
      <description>MCP blog. Our take: Clients must now validate the OAuth issuer before redeeming a code, Dynamic Client Registration is deprecated in favour of client metadata documents, and the legacy HTTP+SSE transport starts a year-long offramp. Expect OAuth sign-in prompts for remote servers to change as clients update. Supabase says the new multi round-trip requests let its tools confirm costly or destructive actions first; see our Supabase survey.</description>
    </item>
    <item>
      <title>The Official Ruby SDK for MCP Reaches 1.0</title>
      <link>https://blog.modelcontextprotocol.io/posts/ruby-sdk-1-0/</link>
      <guid isPermaLink="false">axismcp-news-2026-07-27-the-official-ruby-sdk-for-mcp-reaches-1-0</guid>
      <pubDate>Mon, 27 Jul 2026 12:00:00 GMT</pubDate>
      <category>Spec and SDKs</category>
      <description>MCP blog. Our take: The mcp gem's 1.0 passes every server and client conformance scenario and supports stdio and Streamable HTTP. It implements the 2025-06-18 and 2025-11-25 revisions; 2026-07-28 support is in progress. Relevant if you build servers rather than install them.</description>
    </item>
    <item>
      <title>Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)</title>
      <link>https://glama.ai/blog/2026-07-25-who-is-calling-mcp-hosts-are-an-identity-blind-spot-and-the-spec-knows-it</link>
      <guid isPermaLink="false">axismcp-news-2026-07-25-who-s-calling-mcp-hosts-are-an-identity-blind-spot-and-the-spec-knows-it</guid>
      <pubDate>Sat, 25 Jul 2026 12:00:00 GMT</pubDate>
      <category>Advisory</category>
      <description>Glama blog. Our take: The common default is a long-lived, broad-scope personal access token pasted into an environment variable, which the article calls over-privileged and untraceable. Token-based servers in our directory, such as GitHub's, work that way, so scope the token narrowly and give it an expiry. Remote servers that sign you in with OAuth avoid the standing key.</description>
    </item>
    <item>
      <title>Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)</title>
      <link>https://glama.ai/blog/2026-06-30-why-mcp-servers-need-execution-sandboxing-and-why-your-current-stack-is-not-enough</link>
      <guid isPermaLink="false">axismcp-news-2026-06-30-why-mcp-servers-need-execution-sandboxing-and-why-your-current-stack-isn-t-enough</guid>
      <pubDate>Tue, 30 Jun 2026 12:00:00 GMT</pubDate>
      <category>Advisory</category>
      <description>Glama blog. Our take: It cites three incidents, including CVE-2025-6514, and argues Docker and IAM alone don't bound what an agent-driven tool does; its advice is to match the sandbox to the workload and keep network egress rules beneath it. Before running a local server, check which folders and networks it can reach. The Filesystem server, for example, only gets the directories you list.</description>
    </item>
  </channel>
</rss>
