Waiting Is Not a Tool Call: Making an MCP Server's Shell Event-Driven

<p>Two days into a batch classification run, I opened the log and saw line one again. Not a continuation of line one — the same line one, with a new timestamp and a new process

<p>An inference server stopped generating tokens while more than 200 GiB of host memory remained available. Its health endpoint kept returning HTTP 200. The first visible

<p>The bug report was: "the browser tools are gone."</p> <p>I'd been running the same Claude Code session for an hour, calling <code>safari_navigate</code>,

<p>MCP is one of the most thoughtfully scoped technical standards to come out of the AI ecosystem. That scope is intentional. Understanding exactly where MCP stops is the

<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1wxfdt9/qwen38_27b_1_x_r9700_262k_context_half_a_million/"> <img alt="Qwen3.8 27B | 1 x R9700: 262K context,

<p>In January and February 2026, security researchers filed <a href="https://www.heyuan110.com/posts/ai/2026-03-10-mcp-security-2026/" rel="noopener noreferrer">30 CVEs against

<p>Nearly every guide to setting up an MCP server tells you to do this:<br /> </p> <div class="highlight js-code-highlight"> <pre class="highlight json"><code><span

<p>MCP feels easy until it isn't. The first time you wire up a stdio server and call a tool from a Claude Agent SDK loop, the whole thing fits on a slide. Then you put it in front