alexar76/warden

Overview

WARDEN is a security firewall for MCP servers. It vets a server's tool definitions against static-scan rules, a signed threat feed, and origin and tool-definition pinning before any tool reaches the model. It is a zero-dependency TypeScript library.

Best for

Best for
Developers building MCP servers who need a lightweight pre-model security check

Use cases

• Block malicious or risky MCP tools before model access
• Enforce allowed tool origins and pinned definitions
• Add a static security scan step to MCP server deployment

How to use

Install

npx -y @aimarket/warden            # bin: warden-mcp

Tools exposed

• vet_mcp_server
• static_scan_tools
• classify_sensitive_tools
• check_egress_url
• canonicalize_json
• list_scan_rules

Tested with

Claude Desktop, Cursor

Notes

WARDEN is a security firewall for MCP servers. It vets a server’s tool definitions against static-scan rules, a signed threat feed, and origin and tool-definition pinning before any tool reaches the model. It is a zero-dependency TypeScript library.

0 stars on GitHub. Last updated 2026-09-05. Licensed MIT.

Use cases

• Block malicious or risky MCP tools before model access
• Enforce allowed tool origins and pinned definitions
• Add a static security scan step to MCP server deployment

Pros

• Zero-dependency TypeScript library
• Layers static scans, signed feeds, and pinning
• Runs before any tool reaches the model

Cons

• Zero stars indicate no community validation yet
• Only vets tool definitions, not runtime tool behavior
• Requires keeping rules and threat feed current

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.