MCP pages should read like API documentation: capability, permission, data scope, and revocation

MCP server, MCP tools, agent tools, and MCP security searches show that developers have moved from concept curiosity to safe integration questions.

Back to this issue Open original source
MCP pages should read like API documentation: capability, permission, data scope, and revocation
Radar take

If an AI tool supports MCP, the page should state what it can read, what it can change, which authorization it needs, where logs live, and how to disable it.

MCP is a distribution layer and a trust layer. Clear permissions make team adoption easier.

MCPAgent ToolsDeveloper Workflow