Aegis MCP
未认领Runtime enforcement. Loads policy into server memory, validates every agent action before execution. Blocked actions get a specific reason and an override protocol. Zero token overhead — agents never load governance files.
安装
aegis-mcp --project . --role backend更多AI & ML服务器
浏览完整目录安全概况
已认领和已认证的服务器每周接受一次静态扫描，显示代码能触及的范围（外部服务、环境变量、shell 命令、代理配置目录）以及带有已知公告的依赖。认领此列表即可获得。 安全概况的工作原理
9 个工具中显示 9 个
文档中列出的工具 (9)
内容来自项目文档。服务器握手不会验证每个工具的说明或行为。
aegis_check_permissions
Pre-check if an operation is allowed
aegis_complete_task
Run quality gates and close construction mode if active
aegis_delete_file
Governed delete with path validation
aegis_execute
Governed command execution
aegis_policy_summary
Role boundaries and governance summary (or available roles in universal mode)
aegis_read_file
Governed read with path validation
aegis_request_override
Execute a blocked action after human confirmation
aegis_select_role
Select a role (including construction) in universal mode
aegis_write_file
Governed write with path + content validation