Articoli su security
5 articoli
Segui i nuovi articoli in qualsiasi lettore di feed con il feed Atom.
Gli articoli sono in inglese.
-
MCP Tool Poisoning: Reading Schemas, Enums and Defaults
MCP tool poisoning goes beyond the description field: parameter descriptions, enums, defaults and invisible characters. How to dump a tool list and read it.
-
MCP Tool Annotations: What readOnlyHint and destructiveHint Mean
MCP tool annotations explained: readOnlyHint, destructiveHint, idempotentHint, openWorldHint, their defaults, SDK syntax and why clients treat them as hints.
-
MCP Authorization in Practice: Metadata, Client IDs and Sign-In
How MCP authorization works on the wire: the 401 challenge, protected resource metadata, Client ID Metadata Documents, PKCE, scopes and the client side.
-
Is This MCP Server Safe? How to Check Before You Connect
A practical MCP server security check: tool poisoning, prompt injection, scopes, read-only hints, vendor verification, npm supply chain, and a checklist.
-
Remote vs Local MCP Servers: OAuth, API Keys and stdio
When to run an MCP server locally over stdio and when to use a remote one with OAuth or an API key: security, privacy, performance and enterprise tradeoffs.