AI Agent MCP Server with Input Validation and Other Security Features

Guardrails MCP Server

The AI agent security guardrails MCP server provides security functions such as input validation, prompt injection detection, PII desensitization, output filtering, policy execution, rate limiting, and audit logging.

Security Artificial intelligence chatbots #AI security #Input validation #Audit logging #Policy execution .Rust

rating : 2 points

downloads : 4.9K

update time : 2026-03-12

Open Site

What is Guardrails MCP Server?

Guardrails MCP Server is a security protection middleware specifically designed for AI applications. It acts like an intelligent firewall, performing security checks before user requests reach the AI model and filtering content before AI responses are returned to users. It can detect and block malicious prompts, protect sensitive information, enforce security policies, and record all operations for auditing.

How to use Guardrails MCP Server?

You can integrate Guardrails into your AI application through simple configuration. It runs as an independent MCP server, receiving requests from the AI application, forwarding them to the AI model after performing security checks, and then returning the filtered responses to the application. The entire process is transparent to users, and there is no need to modify the existing business logic.

Applicable scenarios

It is suitable for all application scenarios that need to interact with AI models, especially AI applications that handle sensitive data, serve the public, or require compliance auditing. For example: customer service chatbots, content generation tools, code assistants, data analysis platforms, etc.

Main features

Input validation and protection

Detect and block malicious prompt injection attacks in real - time, including 12 common attack patterns such as jailbreak attacks, system prompt overwriting, and DAN mode, to protect the AI model from being manipulated.

Sensitive information protection

Automatically detect and filter personal identity information (PII), including social security numbers, credit card numbers, email addresses, phone numbers, IP addresses, etc., to prevent data breaches.

Malicious code interception

Block requests containing dangerous code, such as eval, exec, and subprocess calls, to prevent code injection attacks.

Policy enforcement engine

Role - based access control (RBAC), resource quota management, maintenance window control, URL whitelisting, etc., to ensure compliant operations.

Intelligent rate limiting

Request frequency control based on a sliding window, allowing different request rate limits to be configured according to users, to prevent abuse and DDoS attacks.

Comprehensive audit logging

Record all security events and operations, support querying by type, user, and time range, and provide complete traceability.

Runtime configuration update

Dynamically update security policies and configurations without restarting the service, and quickly respond to changes in security requirements.

Advantages

Multi - layer protection: A complete security chain from input validation to output filtering

Easy to integrate: Standard MCP protocol, compatible with various AI frameworks

High performance: Lightweight design, with minimal impact on application performance

Flexible configuration: Support for dynamically adjusting security policies

Comprehensive auditing: Complete log recording to meet compliance requirements

Active protection: Not only detect but also actively block attacks

Limitations

Requires additional deployment: Runs as an independent service, increasing system complexity

Rules need to be maintained: Security rules need to be updated regularly to address new threats

Possible false positives: Strict filtering rules may occasionally misjudge normal requests

Learning cost: Need to understand the MCP protocol and basic configuration

Relies on regular expressions: Some detections are based on regular matching and may be bypassed

How to use

Install dependencies

Ensure that Node.js 18 or a higher version is installed on the system, and then install the project dependency packages.

Configure the MCP client

Configure the MCP client in your AI application and add the connection information of the Guardrails server.

Start the service

Start the Guardrails MCP server, and it will automatically start listening for and processing security requests.

Integrate into the application

Modify the code of your AI application and send all AI requests to Guardrails for security checks first.

Monitor and adjust

Monitor security events through audit logs and adjust security policies and configurations as needed.

Usage examples

Protect customer service chatbots

The customer service chatbot of an e - commerce company needs to handle user inquiries but may receive malicious prompts or leak user privacy information.

Security protection for code generation assistants

Developers use AI assistants to generate code but need to prevent the generation of malicious code or the leakage of API keys.

Content review and compliance

A content creation platform uses AI to generate articles and needs to ensure that no违规 content is generated and user privacy is protected.

Frequently Asked Questions

Will Guardrails affect the response speed of AI applications?

What if Guardrails misjudges a normal request?

Can Guardrails detect all types of attacks?

Is programming knowledge required to use it?

Which AI models and frameworks does Guardrails support?

How to handle false positives and false negatives?

Related resources

Official documentation

Official documentation and specifications of the Model Context Protocol

GitHub repository

Project source code and issue tracking

Security best practices guide

Guide for secure deployment and configuration of AI applications

Community forum

User communication, problem discussion, and case sharing

Configuration example library

Configuration examples and templates for various usage scenarios

🚀 guardrails-mcp-server

MCP server for AI agent security guardrails, offering input validation, prompt injection detection, PII redaction, output filtering, policy enforcement, rate limiting, and comprehensive audit logging.

🚀 Quick Start

The guardrails-mcp-server is a MCP server designed to provide security guardrails for AI agents. It offers a series of security features to ensure the safety and compliance of AI interactions.

✨ Features

Input Validation: Validate and sanitize incoming requests through all guardrail checks.
Output Filtering: Filter and redact sensitive data (PII, secrets, credentials) from responses.
Policy Enforcement: Evaluate requests against security policies (RBAC, resource access, quotas).
Audit Logging: Query the audit log with filtering by type, user, time range.
Statistics Monitoring: Get engine statistics including active users, block rate, request counts.
Runtime Configuration: Update guardrail configuration at runtime.

📦 Installation

npm install

📚 Documentation

Tools

Property	Details
`validate_input`	Validate and sanitize incoming requests through all guardrail checks
`filter_output`	Filter and redact sensitive data (PII, secrets, credentials) from responses
`check_policy`	Evaluate a request against security policies (RBAC, resource access, quotas)
`get_audit_logs`	Query the audit log with filtering by type, user, time range
`get_stats`	Get engine statistics including active users, block rate, request counts
`update_config`	Update guardrail configuration at runtime

Security Features

Prompt Injection Detection: Utilizes 12 regex patterns for jailbreak, DAN mode, and system prompt override.
PII Detection and Redaction: Covers SSN, credit card, email, phone, IP, API keys, AWS keys, JWT, passwords, private keys, and connection strings.
Malicious Code Blocking: Blocks eval, exec, subprocess, child_process, and shell injection.
Policy Engine: Blocks sensitive paths, dangerous tools, unauthenticated destructive ops, maintains a URL allowlist, and sets maintenance windows.
Rate Limiting: Implements per - user sliding window (configurable requests/minute).
Audit Logging: Records timestamped events with metrics, supports log rotation and external handlers.

Configuration

{
  "mcpServers": {
    "guardrails": {
      "type": "stdio",
      "command": "node",
      "args": ["/path/to/guardrails-mcp-server/index.js"]
    }
  }
}

Project Structure

src/engine/GuardrailsEngine.js    # Core orchestration
src/validators/InputValidator.js   # Prompt injection and PII detection
src/filters/OutputFilter.js        # Redaction and harmful content blocking
src/policies/PolicyEngine.js       # RBAC, quotas, maintenance windows
src/audit/AuditLogger.js           # Event logging and metrics