AIM MCP Server is a Model Context Protocol (MCP) server focused on providing security protection for MCP and AI agents. Through AI-driven security analysis and security instruction tools, it provides security guidance, content analysis, and warning instructions when AI agents interact with various MCPs and external services.
Its main capabilities are divided into three major categories:
-
Tools (6 total)
-
AI Safety Guard: Provides contextual security instructions and precautions before an AI agent interacts with other MCPs, with context-aware guidance, operation-specific warnings, and red flag detection.
-
Text Guard Analysis: Uses the AIM Intelligence API to analyze whether text content contains harmful or inappropriate content, supporting real-time analysis, harmful content detection, and detailed JSON results.
-
Security Prompt Enhancement: Adds security instructions to user prompts, supporting multi-level enhancement, threat analysis, and social engineering protection.
-
Prompt Injection Detector: Complies with OWASP LLM01:2025, can detect more than 15 injection patterns (instruction override, role manipulation, jailbreak attempts), provides a 0-100 risk score and severity assessment, configurable sensitivity levels, and detailed threat reports.
-
Credential Scanner: Scans text for exposed credentials, including API keys, passwords, tokens, and SSH keys, supports more than 50 credential patterns (AWS, GitHub, Google, OpenAI, Stripe, JWT, SSH keys), automatically masks credentials, assesses risk levels, performs platform-specific detection, and provides actionable security recommendations.
-
URL Security Validator: Verifies whether URLs have phishing, malware, and security issues, includes more than 10 security checks (protocol, top-level domain, IP address, homoglyph attacks), detects phishing domains, identifies shortened links, detects suspicious parameters, and enforces HTTPS.
-
Resources (9 total) Access read-only security documentation and policies through URI schemes. The security checklist can be accessed via security-checklist://[type] Access, including database (database operations), email (email operations), slack (chat/messaging operations), file (file operations), web (Web requests), and general (general MCP operations). Each checklist includes pre-operation checks, in-operation guidelines, post-operation verification, and red flags indicating when to abort the operation. Security policies can be accessed via security-policy://[type], including data-classification (data classification levels and handling requirements), access-control (access control principles and authentication requirements), and incident-response (incident response process and severity levels).
III. Prompts (2 total)
- Security Review: a comprehensive security review workflow for code, data, or configuration, including credential scanning, prompt injection detection, security checklist review, policy compliance review, threat analysis, risk assessment and recommendations, and a table summarizing by severity.
- Threat Analysis: analyzes potential security threats based on the STRIDE methodology, including asset identification, STRIDE threat modeling (spoofing, tampering, repudiation, information disclosure, denial of service, elevation of privilege), risk assessment (likelihood x impact), attack vector analysis, control gap identification, mitigation strategies, compliance considerations, and incident response planning.
Usage: can be automatically installed for Claude Desktop via Smithery, and also supports NPX (recommended), global installation, and local installation, then added to the MCP client configuration. The project is built using TypeScript and Zod validation, emphasizes being fast and lightweight, can integrate with any MCP-compatible AI assistant, and connects to the AIM Intelligence API for advanced analysis. The project uses the ISC license and includes documentation such as the MCP component guide, GitHub Wiki, and MCP specification.
Use cases: for developers and teams that need to provide security protection for interactions between AI agents and MCP, such as performing security checks before interacting with MCPs such as databases, email, Slack, files, and the Web, detecting prompt injection, scanning for exposed credentials, verifying URL security, and performing security reviews and STRIDE threat modeling.
Limitations: specific feature effectiveness, fees, and pricing information are not yet provided.