MCP SERVER SCANNER

MCP servers are code.
Code can be weaponized.

MCP servers run with full system access. A single malicious tool can exfiltrate data, execute commands, or poison your AI's context. Scandar analyzes the actual source code before it gets anywhere near your machine.

Scan an MCP ServerRead the Docs
WHAT IT DETECTS

8 threat categories for server-side code.

Tool Poisoning
Detects MCP tools with hidden instructions, deceptive descriptions, or shadow functionality.
Hardcoded Secrets
Catches API keys, tokens, passwords, and credentials embedded directly in source code.
Unsafe Command Execution
Flags shell exec, child_process, subprocess calls, and arbitrary code execution patterns.
Unrestricted File Access
Identifies path traversal, unrestricted reads/writes, and directory enumeration.
External Data Transmission
Catches unauthorized HTTP requests, webhook calls, and data exfiltration channels.
Supply Chain
Detects suspicious dependencies, dynamic imports, and post-install script attacks.
Transport Security
Flags insecure transports, missing TLS, and cleartext credential transmission.
Credential Harvesting
Spots phishing patterns, fake auth prompts, and credential interception techniques.
SUPPORTED LANGUAGES

Scan servers written in any major language.

TypeScript / JavaScript.ts .js .tsx .jsx
Python.py
Go.go
Rust.rs
Java.java
ALSO ON SCANDAR
Skill Scanner
Pattern-based and LLM-powered analysis for AI skill files. Detect prompt injection, credential theft, and more.
AI Fix
Select threats and let Claude automatically rewrite your code with threats removed. Review the diff, then download.

Scan an MCP Server

Drop a server file or paste the source code. Results in seconds.

Scan an MCP Server
ENTERPRISE

Need fleet-wide AI security?

Scandar Overwatch gives you real-time visibility into every agent in your organization — policies, compliance reports, alert routing, and kill chain detection. Self-serve setup in 25 minutes.

Explore Overwatch →