MCP-S · v1.0 · 2026 8 ENGINES · 81 RULES · <10s SCAN

Scan Every
MCP Server
in Seconds.

8 detection engines analyze your Model Context Protocol servers for secrets, prompt injection, supply chain risks, data exfiltration paths, schema rug pulls, and over-privileged tools — all in one scan.

Live Scan Result gitmcp.io/docs

gitmcp.io/docs

2 servers · 5 tools · 5.9s

0 CRIT 12 HIGH 0 LOW

HIGHASH-008SSRF surface: url in fetch_generic_url

HIGHASH-007Cmd injection: query in search_docs

HIGHASH-005SQL injection surface: repo_name

HIGHCPA-002Broadly capable tool: match_libs

HIGHASH-009Template injection: owner param

HIGHASH-008SSRF: search_generic_code query

8 engines · 5.9s scan-20260224

How It Works

Three ways to scan.
Same 8 engines every time.

Point at any MCP server

Paste a URL, drop in your claude_desktop_config.json, or upload a config file. Supports SSE and Streamable HTTP transports.

8 engines fire in parallel

SDE, STA, SIM, PPV, COI, DFP, CPA, and ASH each analyze tool definitions, schemas, packages, and data flows simultaneously — no waiting.

ENABLES wires attack chains

The ENABLES engine links related findings into multi-step attack chains — showing how a malicious package enables credential theft enables data exfiltration.

Score, remediate, repeat

Security score, severity breakdown, knowledge graph, and remediation steps — all persisted to scan history. Compare results over time.

terminal — praqtor-mcp CLI

# Install

$pip install praqtor-mcp

# Scan a live MCP server by URL

$praqtor-mcp scan-url \

https://gitmcp.io/docs \

--api-key pmcp-...

Security Score: 62 / 100

Findings: 12 total

Critical 0 Medium 0

High 12 Low 0

# Or scan a config file

$praqtor-mcp scan config.json \

--api-key pmcp-...

# SARIF output for GitHub Actions

$praqtor-mcp scan-url https://... \

--format sarif > results.sarif

✓ Results stored · dashboard ready

Detection Engines 8 of 8 active

Code	Engine	What It Finds	Status
SDE	Secrets Scanner	API keys, tokens, passwords, and credentials hardcoded in MCP server configurations and environment variables.	LIVE
STA	Prompt Threat Detector	Hidden instructions, prompt injection, and jailbreak attempts embedded in tool descriptions and schema annotations.	LIVE
SIM	Schema Change Monitor	Rug pull attacks — tool definitions that silently mutate between scans to include malicious directives or expanded permissions.	LIVE
PPV	Package Risk Checker	Malicious npm and PyPI packages, typosquatting, unpinned dependency versions, and supply chain compromise indicators.	LIVE
COI	Cross-Server Threat Detector	Cross-origin influence attacks where one MCP server's tool manipulates, shadows, or overrides another server's behavior.	LIVE
DFP	Data Exfiltration Tracer	Data movement from sensitive sources to external sinks — mapping hidden exfiltration paths across tools and servers.	LIVE
CPA	Permission Overreach Scanner	Over-privileged tools requesting more capabilities than their stated purpose requires — scored per tool with alignment metrics.	LIVE
ASH	Input Attack Surface Analyzer	SQL injection, SSRF, command injection, path traversal, and template injection surfaces in tool parameter schemas.	LIVE

Screenshots Real data · No mocks

Dashboard

Security Score at a Glance

Live score, finding severity breakdown, active alerts, and detection coverage — all populated after every scan.

Findings

Every Threat, Fully Explained

Filter by severity and engine. Each finding links to its policy rule with remediation steps and OWASP mapping.

Knowledge Graph

Attack Chain Visualization

Interactive graph connecting servers, tools, and findings into multi-step attack chains with blast radius scoring.

Rule Catalog

81 Policy Rules, Fully Documented

Detection patterns, remediation steps, OWASP/CWE mappings, config examples, and per-rule enable/disable for every rule.

Scan Every MCP Server in Seconds.

Three ways to scan.Same 8 engines every time.

Every MCP server is anattack surface.

Scan Every
MCP Server
in Seconds.

Three ways to scan.
Same 8 engines every time.

Every MCP server is an
attack surface.