Guardrails

A local MCP server that adds a security guardrail layer to AI agents: input validation, prompt-injection detection, PII redaction, policy checks and audit logs.

What you can do

Validate and sanitize incoming requests through all guardrail checks, including prompt-injection detection and malicious code blocking. Filter and redact sensitive data such as PII, secrets and credentials from responses. Evaluate requests against security policies covering RBAC, resource access and quotas. Query audit logs by type, user and time range, and read engine statistics such as active users, block rate and request counts. Update guardrail configuration at runtime.

When to use it

Useful when an assistant or agent handles untrusted input and you want a local checkpoint that screens requests, redacts sensitive output and keeps an audit trail. Suited to setups that need policy enforcement, rate limiting or a record of blocked requests.

Requirements

Runs as a local stdio process on the user's machine, installed from the npm package guardrails-mcp-server and started with Node.js. The README shows a manual configuration pointing at the server's index.js path. No authentication, environment variables or headers are declared.