Heimdall — MCP Security Scanner

Name: Heimdall — MCP Security Scanner
Author: caglarbozkurt

@caglarbozkurt

リポジトリホームページ

基本情報

カテゴリ

開発者ツール

ライセンス

MIT

トランスポート

stdio

作者

caglarbozkurt

投稿者

Çağlar Bozkurt

Heimdall — MCP Security Scanner について

Security scanner for MCP servers — vet an MCP before you wire it into an agent. Detects prompt-injection, credential exfiltration (via taint analysis), RCE, and supply-chain risks, and catches cross-server exfil chains no single server reveals. Zero-dependency local CLI, SARIF ou

#cli #security #mcp #compliance #security-scanner #sarif #ai-agents #sast #supply-chain-security #prompt-injection #model-context-protocol #agent-security

Heimdall — MCP Security Scanner の設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "heimdall": {
      "command": "npx",
      "args": [
        "-y",
        "--package",
        "mcp-heimdall-scan",
        "heimdall-mcp"
      ]
    }
  }
}

Heimdall — MCP Security Scanner のツール

Statically vet an MCP server (or a whole MCP client config) BEFORE trusting or installing it. Checks prompt-injection / tool-poisoning, risky capabilities (fs/network/shell/eval/credentials), proven data-exfiltration paths, dependency CVEs, provenance, and multi-server composition. Returns a pass/warn/fail verdict with cited evidence. Static analysis only — downloads the package but does NOT execute it. Use it when the user is about to add/connect an MCP server.

概要

What is Heimdall — Mcp Security Scanner?

Heimdall is a local, pre-flight security scanner for Model Context Protocol (MCP) servers. It vets a single server or an entire agent configuration before an agent trusts it, without running the server by default, and requires no account or backend.

How to use Heimdall — Mcp Security Scanner?

Use the CLI via npx mcp-heimdall-scan <target> or install globally. Targets include npm/PyPI packages, local directories, git repositories, a tools/list dump, or an MCP client config file (e.g., claude_desktop_config.json). Options like --policy, --online, --json, and --sarif customize the scan. It also ships as a GitHub Action and as an MCP server itself.

Key features of Heimdall — Mcp Security Scanner

Static analysis without executing server code.
Detects injection, capability, exfiltration, provenance, and CVEs.
Proven data-flow paths for JS/TS (file:line evidence).
Audits whole agent configs for cross-server exfiltration chains.
Policy-based verdicts (default, strict, custom waivers with expiry).
Runs entirely locally, no account or backend needed.

Use cases of Heimdall — Mcp Security Scanner

Vet a server package before installing it in your agent.
Audit your entire MCP client config (e.g., claude_desktop_config.json) for cross-server risks.
Gate pull requests in CI using the GitHub Action to fail on risky servers.
Use the MCP server mode to let your agent scan a server before connecting to it.
Run behavioral validation (heimdall validate) in a disposable VM to check static findings against runtime behavior.

FAQ from Heimdall — Mcp

「開発者ツール」の他のコンテンツ

FARPY

mangomunchr

Verified distributed compute for humans, applications, and agents. Submit Blender, Octane, and video workloads, route them across qualified GPUs, track asynchronous execution, download finished artifacts, and verify FARP

LastPing

Know when your AI agents, cron jobs and CI pipelines fail, stall or go quiet. Your agent sets up monitors from a plain description, tells you what is broken and why, reads run history and traces, and writes notes on inci

Everframe

ScriptX

Read Everframe bug reports, crashes and tickets from your AI agent: annotated screenshots, console logs, network requests, breadcrumbs, stack traces and device context from web, iOS, Android and TV apps. Eight read-only

Vibgrate MCP

Vibgrate

Query your team's drift, vulnerability, and migration data from any AI assistant. Vibgrate MCP connects Cursor, Claude, ChatGPT, Windsurf, or VS Code to Vibgrate Cloud: 51 tools for DriftScores, CVEs and EOL runtimes, up

Modelglass

Modelglass turns "which model should I use?" into a lookup instead of a guess. Seven tools give your agent live pricing across image, language, video, and audio models. Sourced and versioned, not scraped once and forgott