Skip to main content
Glama
kartik-modi

fast-mcp-gatekeeper

by kartik-modi

๐Ÿ›ก๏ธ fast-mcp-gatekeeper

The Zero-Latency Security & Policy Proxy for Model Context Protocol (MCP)
Powered by TypeSafe AI's Jev System 1 Decision Model.

npm version License: MIT Powered by Jev

As coding assistants (Cursor, Claude Desktop, Antigravity) gain access to local tools via the Model Context Protocol (MCP), granting unconstrained access to your filesystem, terminal, and database is risky.

Existing security checks run through heavy LLMs add 2,000โ€“3,000ms of latency to every single tool call, crippling agent responsiveness.

fast-mcp-gatekeeper acts as an ultra-fast stdio proxy sitting between your MCP client and any MCP server. It inspects all JSON-RPC tools/call packets in <50 milliseconds, detecting path traversals, command injections, and credential leaks without slowing down your AI workflow.


โœจ Features

  • โšก Sub-50ms Overhead: Jev evaluates tool arguments in a single non-autoregressive pass.

  • ๐Ÿ›‘ Automatic Interception: Returns standard MCP JSON-RPC -32001 error codes before the downstream tool can ever execute.

  • ๐Ÿ”‘ Automatic Token Sanitization: Redacts leaked API keys (sk-..., Bearer tokens, passwords) on the fly with the redact policy.

  • ๐Ÿ”Œ Universal Compatibility: Works with any stdio-based MCP server (e.g. @modelcontextprotocol/server-filesystem, Postgres, Git, GitHub).


Related MCP server: MCP Security Gateway

๐Ÿš€ Quick Setup with Claude Desktop or Cursor

In your claude_desktop_config.json or Cursor MCP settings, simply prepend fast-mcp-gatekeeper -- to your server command:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "fast-mcp-gatekeeper",
        "--",
        "npx",
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/Users/username/workspace"
      ]
    }
  }
}

๐Ÿงช CLI Inspection Demo

You can inspect any simulated tool call directly from the terminal:

npx fast-mcp-gatekeeper --inspect '{"name": "read_file", "arguments": {"path": "/etc/passwd"}}'

Output:

{
  "toolName": "read_file",
  "isMalicious": true,
  "maliciousProbability": 0.99,
  "threatLevel": "critical",
  "policy": "block",
  "reason": "BLOCKED: High-risk security policy violation detected in tool 'read_file'.",
  "latencyMs": 41
}

๐Ÿ“„ License

MIT ยฉ Kartik Modi

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    C
    maintenance
    Enables secure interoperability between LLM agents and MCP tool servers by sanitizing requests and responses, masking sensitive tokens, detecting PII, and performing server reputation scans.
    MIT
  • F
    license
    Not graded
    quality
    C
    maintenance
    Enables safe use of any MCP server by proxying and live-scanning all tool requests and responses, blocking or redacting poison descriptions, indirect prompt injection, malicious arguments, and unauthorized destinations.
    -
  • A
    license
    Not graded
    quality
    B
    maintenance
    Proxies MCP traffic between a client and a downstream server to enforce runtime policies on tool declarations, call arguments, and results, including allowlisting, sandboxing, secret and egress controls, and injection detection. It also includes a deterministic benchmark for measuring which security controls stop which attacks.
    MIT
  • A
    license
    Not graded
    quality
    F
    maintenance
    Enables transparent security for any MCP server by intercepting tool calls, blocking prompt injection attempts, masking PII in responses, and writing immutable audit logs.
    302 npm
    MIT