Urgent.News

What's breaking now, across thousands of outlets.

Tech

Shelf Watch #3: What good looks like on a hosted enterprise MCP server

Originally published in the mcp-ship-check repo, home of the free, offline scanner this post describes. Update (2026-11-11). Since this post first ran, we've finished a static code audit of 300 MCP servers : 10 had a verified code-level vulnerability. One is fixed and published ( atlas-mcp-server , GHSA-w3gh-28jp-2hmv ); the rest are being disclosed privately to their maintainers, and we name a…

The Shelf Watch #3 report outlines best practices for securing hosted enterprise MCP servers, especially those serving multiple tenants. The key findings include four essential controls for a well-guarded server:

1. Login scopes mirroring existing product roles. Clients authenticate using OAuth 2.1 and PKCE, with token scopes mapping to the product's role-based access. This ensures the token only grants the permissions the user already has, preventing over-sharing.

2. Small blast radius by default. Requests per token are capped, typically at 100 per minute. Each call targets a single resource with write tools scoped to a single subject like a device or user. Default validation checks prevent unintentional broad actions.

3. Human approval on every write. Write tools pause for synchronous, in-chat human approval before execution. If approval cannot be obtained, the request fails closed instead of proceeding blindly. Tools should be marked with read/write hints to guide this process.

4. Sensitive tools off the external surface. Money, entitlements, billing, and similar critical functions are not exposed to external MCP clients. These operations remain on internal surfaces accessible only to the vendor's own tools and UI.

The report emphasizes that these controls can be implemented by a small team within a two-person capacity, using existing infrastructure and tools without requiring an enterprise budget. By applying these practices, hosted MCP servers can achieve a higher level of security and prevent unauthorized actions that could compromise multiple tenants.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in Tech

Verifying Remediation for CVE-2026-79863

Verifying Remediation for CVE-2026-79863 A successful software upload is not proof that CVE-2026-79863 is fixed. The device may have rolled back, booted from another partition, retained an older image…

Can You Decrypt a Bcrypt Hash? No, and Here Is What Works Instead

Every week someone asks how to "decrypt" a bcrypt hash. Here is the short answer, and what to do instead. Originally published on BcryptGenerator.org . No, a bcrypt hash can't be decrypted.

  • Bcrypt is a one-way hash function, not decryptable
  • Password recovery requires brute-force guessing, slow process
  • Bcrypt stores passwords securely, resistant to cracking

We Turned Our GLP-1 Price Checks Into a Public Dataset and Live MCP Feed

Comparing GLP-1 programs takes more digging than it should. A provider may advertise one monthly figure, then handle consultation fees, membership fees, dose changes, and billing periods somewhere…

  • GLP-1 price comparisons turned into public dataset
  • Live MCP feed allows developer access to pricing info
  • Dataset built from price checks between Oct 15 and Nov 2, 2026

WooCommerce to Zapier Not Triggering? Check These 5 Things

Disclosure: I'm building Webhook Disable Monitor, a free WooCommerce plugin in pre-release (not published, nothing for sale). It is only relevant to check #1 and I mention it there.

  • Verify WooCommerce data sending setup: official extension or self-created webhook
  • Check if webhook is active; deactivate webhooks after five failed deliveries
  • Ensure background queue (Action Scheduler or WP-Cron) is running for data delivery

More from Friday 13 November →