ai-safety-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@ai-safety-mcp Scan this text for policy issues: "I will hack your account.""
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
@mhdd_24/ai-safety-mcp
MCP server for Evaluate unsafe/problematic model outputs.
Same architecture as @mhdd_24/sublime-mcp.
Full documentation: docs/WIKI.md
How it works (30 seconds)
You (chat) → MCP client → ai-safety-mcp → AI Safety APIs / CLIs / local toolsRelated MCP server: agentguard
Prerequisites
Requirement | Notes |
Node.js 18+ | ESM TypeScript MCP server |
Credentials / CLIs | See environment variables below |
Install
Option A — npm (after publish)
npm install -g @mhdd_24/ai-safety-mcpOption B — npx
npx @mhdd_24/ai-safety-mcpOption C — clone and build
git clone https://github.com/Mhdd-24/AI-Safety-MCP.git
cd AI-Safety-MCP
npm install
npm run build
node dist/index.jsConfigure Cursor
Edit ~/.cursor/mcp.json:
{
"mcpServers": {
"aisafety": {
"command": "npx",
"args": ["-y", "@mhdd_24/ai-safety-mcp"],
"env": {
"_": "optional"
}
}
}
}Local development:
{
"command": "node",
"args": ["/absolute/path/to/AI-Safety-MCP/dist/index.js"]
}Environment variables
Variable | Description |
— | No required env |
Tools
Tool | Description |
| Show Evaluate unsafe/problematic model outputs configuration / health. |
| Scan text for policy/safety issues. |
| Generate a test suite outline. |
License
ISC
Available Tools
3 toolsaisafety_scanC
Scan text for policy/safety issues.
Inputs: text (required).
aisafety_statusC
Show Evaluate unsafe/problematic model outputs configuration / health.
Inputs: no named parameters.
aisafety_suiteC
Generate a test suite outline.
Inputs: policy (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 3 tools
Each tool addresses a clearly distinct activity: checking health/configuration, scanning text for issues, and generating a test suite outline. There is no meaningful overlap between these operations.
All tool names share the consistent 'aisafety_' prefix and are concise, but the naming style mixes nouns (status, suite) with an action-like word (scan). 'aisafety_suite' especially omits an explicit verb such as 'generate' or 'create'.
Three tools is a well-scoped count for a focused AI safety utility server. Each tool earns its place and covers a distinct core capability without unnecessary bloat.
The toolset covers core safety workflow needs: health check, scanning, and test suite generation. Minor gaps exist, such as the absence of a dedicated tool for detailed policy management or report retrieval, but the core surface is not incomplete.
Related MCP Connectors
Scan text, documents, websites, and MCP metadata for prompt injection and sensitive-data risks.
Scan an AI agent's instructions and tools for behavior risks and generate probes that test them.
LLM agent security scan: injection, jailbreak, tool abuse probes plus guardrail audit.
Scan configs, files, or text for leaked secrets and obvious misconfigurations. Nothing stored.
Related MCP Servers
- AlicenseBqualityDmaintenanceEnables testing AI safety classifier robustness against query decomposition, obfuscation, and multi-agent attacks. Provides tools for full evaluation pipelines, query previews, and status checks.46MIT
- AlicenseNot gradedqualityCmaintenanceEnables scanning of AI agent code for security vulnerabilities such as prompt injection, tool abuse, and data exfiltration, directly from MCP-compatible clients like Claude Code.53 PyPI3LGPL 3.0
- AlicenseNot gradedqualityCmaintenanceEnables scanning LLM prompts and responses for prompt injection, jailbreaks, PII leakage, secret leakage, and other malicious content using deterministic rules, returning verdicts and safe redacted text.1MIT
- AlicenseBqualityCmaintenanceEnables input/output policy validation via MCP, including scanning text for policy and safety issues, generating test suite outlines, and checking configuration health.310 npmISC