MCP安全标准

MTF是一个开放安全标准,用于描述和验证MCP服务器包的安全状态,适用于安全扫描、合规评估和采购决策等场景。

  • by NimbleBrainInc
  • Astro
  • 本地部署
  • 1 星标
  • 更新于
在 GitHub 查看 NimbleBrainInc/mpak-trust-framework

使用说明

mpak Trust Framework (MTF)

MTF is an open security standard for describing and verifying the security posture of MCP server bundles.

Status: Draft (v0.1) License: CC BY 4.0

Overview

MCP servers extend AI assistants with powerful capabilities: filesystem access, network requests, database queries, and code execution. This power creates significant security risk. MTF provides a standardized framework for:

  1. Bundle authors to demonstrate security best practices
  2. Registries to enforce minimum security requirements
  3. Consumers to make informed installation decisions
  4. Enterprises to set procurement policies

Compliance Levels

MTF defines four compliance levels, each building on the previous:

Level Name Target Controls
L1 Basic Personal projects, experimentation 6
L2 Standard Team tools, published packages 14
L3 Verified Production, enterprise use 22
L4 Attested Critical infrastructure, regulated industries 25

Security Domains

Controls are organized into five domains:

  • Supply Chain (SC): SBOM, vulnerability scanning, dependency pinning
  • Code Quality (CQ): Secret detection, malicious patterns, static analysis
  • Artifact Integrity (AI): Manifest validation, content hashes, signatures
  • Provenance (PR): Source repository, author identity, build attestation
  • Capability Declaration (CD): Tool declarations, permission scopes

Specification

See MTF-0.1.md for the full specification.

Schemas

JSON schemas for validation:

Implementations

Implementation Language Maintainer
mpak-scanner Python NimbleBrain (reference implementation)

Contributing

MTF is developed in the open. Contributions, feedback, and discussion are welcome.

License

This specification is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).

You are free to share and adapt this material for any purpose, including commercial use, as long as you provide appropriate attribution.