The latest articles on DEV Community by MCP Verified (@mcpverified). https://dev.to/mcpverified https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4201502%2Fb9335ae7-267b-4ebf-bb28-8ced6a39ce4a.png https://dev.to/mcpverified en MCP Verified Fri, 23 Oct 2026 14:10:29 +0000 https://dev.to/mcpverified/what-breaks-in-mcp-servers-results-from-launching-750-registry-entries-h0p https://dev.to/mcpverified/what-breaks-in-mcp-servers-results-from-launching-750-registry-entries-h0p <p><em>Disclosure: I run MCP Verified, a weekly test of servers in the official Model Context Protocol registry. This post is the data. The newsletter link is at the end.</em></p> <h2> What we test </h2> <p>Each week we take a sample of 750 entries from the official MCP registry (300 npm, 150 PyPI, 200 remote streamable-HTTP, 100 remote SSE). We install or connect to each one in a clean sandbox, run the MCP initialize handshake, and call <code>tools/list</code>. Where possible we also make one harmless tool call. We supply no API keys.</p> <h2> Results (run of Oct 19, 2026) </h2> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Type</th> <th>Tested</th> <th>Working</th> <th>Median handshake</th> <th>Median tools</th> </tr> </thead> <tbody> <tr> <td>npm</td> <td>300</td> <td>218 (73%)</td> <td>563 ms</td> <td>14</td> </tr> <tr> <td>PyPI</td> <td>150</td> <td>73 (49%)</td> <td>1,581 ms</td> <td>16</td> </tr> <tr> <td>HTTP</td> <td>200</td> <td>130 (65%)</td> <td>436 ms</td> <td>8.5</td> </tr> <tr> <td>SSE</td> <td>100</td> <td>10 (10%)</td> <td>1,152 ms</td> <td>9</td> </tr> <tr> <td><strong>All</strong></td> <td><strong>750</strong></td> <td><strong>431 (57%)</strong></td> <td></td> <td></td> </tr> </tbody> </table></div> <h2> Failure reasons (319 servers) </h2> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Reason</th> <th>Count</th> </tr> </thead> <tbody> <tr> <td>Needs config/env vars (untested)</td> <td>92</td> </tr> <tr> <td>Remote endpoint dead</td> <td>88</td> </tr> <tr> <td>Needs credentials (untested)</td> <td>54</td> </tr> <tr> <td>Crashed on start</td> <td>47</td> </tr> <tr> <td>Install failed</td> <td>12</td> </tr> <tr> <td>Handshake failed</td> <td>10</td> </tr> <tr> <td>tools/list failed after handshake</td> <td>6</td> </tr> <tr> <td>URL isn't an MCP server</td> <td>6</td> </tr> <tr> <td>Timeout (45 s)</td> <td>4</td> </tr> </tbody> </table></div> <h2> Practical checklist for server authors </h2> <ol> <li>Don't crash when an env var is missing. Finish the handshake and return a clear error from the tool call.</li> <li>If you publish a remote endpoint, monitor it. 69 of 100 SSE entries were dead.</li> <li>Pin dependencies. Install failures were few (12), but each one is a total loss for users.</li> <li>Keep startup fast. Clients often time out. The PyPI median handshake was about 3× npm's.</li> </ol> <h2> Trend </h2> <p>Working count in our runs: 422 (Oct 13), then 427, then 431 (both Oct 19).</p> <h2> Caveats </h2> <p>This is a sample, not the whole registry. "Needs config" doesn't mean broken. Results are from one sandbox environment.</p> <p>Weekly results (free newsletter): <a href="https://buttondown.com/mcpverified" rel="noopener noreferrer">https://buttondown.com/mcpverified</a> · Site: <a href="https://mcpverified.netlify.app" rel="noopener noreferrer">https://mcpverified.netlify.app</a></p> mcp ai opensource testing