Docs/Security/Overview

How Micipi Secures Your MCP Servers

MCP servers are a new and rapidly growing attack surface. Micipi provides automated security scanning and enforcement to keep your servers safe.

The Problem

2026 has seen an explosion of MCP-related security issues. In the first 60 days alone, over 30 CVEs were filed against MCP servers and clients. Independent research has revealed alarming statistics:

30+
CVEs in 60 days
82%
Vulnerable to path traversal
43%
Basic shell injection
71%
No input validation

Most MCP servers are built quickly by individual developers who may not have security expertise. Without proper scanning and validation, these servers become vectors for data exfiltration, remote code execution, and prompt injection attacks.

Real-World Incidents

GitHub MCP Prompt Injection

Attackers embedded hidden instructions in GitHub issue descriptions that manipulated MCP-connected AI agents into exfiltrating private repository data.

WhatsApp MCP Data Exfiltration

A vulnerable WhatsApp MCP server allowed attackers to read and forward private messages through crafted tool calls, compromising user conversations.

Anthropic MCP Inspector RCE

A remote code execution vulnerability was discovered in the official MCP Inspector tool, allowing arbitrary command execution on developer machines.

Memory Poisoning Attacks

Attackers used MCP tool responses to inject false context into AI agent memory systems, causing persistent manipulation of agent behavior across sessions.

How Micipi Protects You

Every server deployed through Micipi undergoes an automated security scan that checks for common vulnerability patterns. Servers receive a score out of 100, and those scoring below 50 are blocked from the marketplace.

Path Traversal Detection
Identifies file system access patterns that could allow attackers to read or write files outside intended directories.
Shell Injection Analysis
Detects unsafe command construction that could allow arbitrary shell command execution through crafted inputs.
Input Validation Verification
Ensures all tool inputs are properly validated with type checking, length limits, and format constraints.
Prompt Injection Defense
Scans for patterns where user-supplied data could manipulate AI agent behavior through embedded instructions.
Resource Exhaustion Prevention
Checks for unbounded loops, unlimited data fetching, and missing timeouts that could cause denial of service.
Data Exfiltration Monitoring
Identifies outbound network calls that could be used to leak sensitive data to external servers.
Authentication and Authorization
Verifies that servers properly authenticate callers and enforce access control on sensitive operations.
Dependency Vulnerability Scanning
Checks imported packages against known vulnerability databases for supply chain attack prevention.

OWASP MCP Top 10

The OWASP Foundation has published a draft Top 10 for MCP security risks. Micipi's scanner addresses each of these categories:

1
Tool Poisoning — Malicious tool descriptions that manipulate AI behavior
2
Excessive Agency — Tools with overly broad permissions and capabilities
3
Insecure Data Handling — Improper handling of sensitive data in tool inputs/outputs
4
Server-Side Request Forgery — Using MCP tools to access internal network resources
5
Prompt Injection via Tools — Tool responses that contain hidden instructions
6
Supply Chain Attacks — Compromised dependencies in MCP server packages
7
Insufficient Logging — Missing audit trails for security-sensitive operations
8
Insecure Configuration — Default settings that expose unnecessary attack surface
9
Resource Exhaustion — Denial of service through unbounded resource consumption
10
Authentication Bypass — Weaknesses in server authentication mechanisms

Deploy with confidence

Every Micipi server is scanned before deployment. Start building secure MCP servers today.

Get Started Free