ToolTrust
Scans MCP servers for prompt injection, supply chain attacks, excessive permissions, and code execution risks. Includes an offline blacklist that catches known-compromised packages like LiteLLM 1.82.7/1.82.8 and Trivy with zero latency.
What is ToolTrust?
ToolTrust is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to scans mcp servers for prompt injection, supply chain attacks, excessive permissions, and code execution risks. includes an offline blacklist that catches known-compromised packages like litellm 1.82.7...
Scans MCP servers for prompt injection, supply chain attacks, excessive permissions, and code execution risks. Includes an offline blacklist that catches known-compromised packages like LiteLLM 1.82.7/1.82.8 and Trivy with zero latency.
This server falls under the Security category on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- Scans MCP servers for prompt injection, supply chain attacks
Use Cases
Maintainer
Works with
Installation
NPM
npx -y tooltrust-mcpManual Installation
npx -y tooltrust-mcpConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use ToolTrust
ToolTrust is a security scanner for MCP servers written in Go that performs static analysis of MCP tool definitions, detecting prompt injection vulnerabilities, supply chain attacks, excessive permissions, and code execution risks across 19 security rule categories. It operates as an MCP server itself (via stdio transport), exposing five tools that let AI assistants audit their own MCP configuration, scan specific servers, analyze raw tool definitions, and look up known-compromised packages via a zero-latency offline blacklist that catches confirmed malicious packages like LiteLLM 1.82.7/1.82.8 and Trivy 0.69.4-0.69.6 without a network call. Security-conscious teams deploying AI agents with multiple MCP servers will find ToolTrust essential for continuous supply chain and permission auditing.
Prerequisites
- Node.js 18 or higher (for npx installation) or Go 1.21+ (for go install)
- An MCP-capable client such as Claude Desktop or Claude Code
- No API keys required — all scanning is performed locally
Install via npx (no installation needed)
Run ToolTrust directly with npx. This downloads and executes the latest version without a global install.
npx -y tooltrust-mcpInstall via Go (alternative)
If you have Go installed, install the binary directly from the source repository.
go install github.com/AgentSafe-AI/tooltrust-scanner/cmd/tooltrust-scanner@latestOne-line install script (macOS/Linux)
Use the official install script for a managed installation on macOS or Linux.
curl -sfL https://raw.githubusercontent.com/AgentSafe-AI/tooltrust-scanner/main/install.sh | bashConfigure Claude Desktop
Add ToolTrust to your Claude Desktop MCP configuration so it launches automatically and can audit the rest of your MCP server setup.
{
"mcpServers": {
"tooltrust": {
"command": "npx",
"args": ["-y", "tooltrust-mcp"]
}
}
}Run a security scan of your MCP configuration
Once connected, ask Claude to audit all configured MCP servers. ToolTrust will analyze tool definitions against its 19 security rules and report grades from A to F.
ToolTrust Examples
Client configuration
Configure Claude Desktop to use ToolTrust as an MCP server for on-demand security scanning.
{
"mcpServers": {
"tooltrust": {
"command": "npx",
"args": ["-y", "tooltrust-mcp"]
}
}
}Prompts to try
Use these prompts in Claude to run security scans on your MCP server setup.
- "Scan all my configured MCP servers for security issues"
- "Check if any of my MCP servers have prompt injection vulnerabilities"
- "Look up LiteLLM 1.82.7 in the vulnerability blacklist"
- "List all the security rules that ToolTrust checks for"
- "Scan the filesystem MCP server and grade its tool definitions"
- "Are any of my installed MCP packages on the known-compromised blacklist?"Troubleshooting ToolTrust
npx command fails or times out on first run
The first run downloads the Go binary. Ensure Node.js 18+ is installed and your network can reach the npm registry. Try 'npm install -g tooltrust-mcp' for a persistent global install to avoid repeated downloads.
tooltrust_scan_config returns no results or errors
The tool looks for the MCP configuration file in the default location for your OS (e.g., ~/Library/Application Support/Claude/claude_desktop_config.json on macOS). Ensure Claude Desktop is installed and has a valid config file with at least one mcpServers entry.
A server receives a low grade (D or F) but seems safe
Use tooltrust_list_rules to see the full list of 19 security rules and which ones triggered. Some rules flag broad permissions or overly-permissive tool descriptions that could be tightened even if not actively exploited. Review each flagged rule's description for specific remediation guidance.
Frequently Asked Questions about ToolTrust
What is ToolTrust?
ToolTrust is a Model Context Protocol (MCP) server that scans mcp servers for prompt injection, supply chain attacks, excessive permissions, and code execution risks. includes an offline blacklist that catches known-compromised packages like litellm 1.82.7/1.82.8 and trivy with zero latency. It connects AI assistants to external tools and data sources through a standardized interface.
How do I install ToolTrust?
Install via npm with the command: npx -y tooltrust-mcp. Then add the server configuration to your AI client's JSON config file (e.g., claude_desktop_config.json or .cursor/mcp.json).
Which AI clients work with ToolTrust?
ToolTrust works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is ToolTrust free to use?
Yes, ToolTrust is open source and available under the MIT license. You can use it freely in both personal and commercial projects.
ToolTrust Alternatives — Similar Security Servers
Looking for alternatives to ToolTrust? Here are other popular security servers you can use with Claude, Cursor, and VS Code.
Casdoor
★ 13.6kAn open-source Agent-first Identity and Access Management (IAM) /LLM MCP & agent gateway and auth server with web UI supporting OpenClaw, MCP, OAuth, OIDC, SAML, CAS, LDAP, SCIM, WebAuthn, TOTP, MFA, Face ID, Google Workspace, Azure AD
ghidraMCP
★ 9.0kAn Model Context Protocol server that enables LLMs to autonomously reverse engineer applications by exposing Ghidra's decompilation and analysis tools. It allows AI agents to list code structures, rename methods, and analyze binaries directly through
HexStrike AI
★ 8.9kHexStrike AI MCP Agents is an advanced MCP server that lets AI agents (Claude, GPT, Copilot, etc.) autonomously run 150+ cybersecurity tools for automated pentesting, vulnerability discovery, bug bounty automation, and security research. Seamlessly b
IDA Pro MCP
★ 8.7kEnables AI-assisted reverse engineering in IDA Pro by providing tools to analyze binaries, decompile functions, manage comments, search patterns, and interact with the IDA database through natural language.
Anthropic Cybersecurity Skills
★ 6.6k754 structured cybersecurity skills for AI agents · Mapped to 5 frameworks: MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND & NIST AI RMF · agentskills.io standard · Works with Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI & 20+ platform
Hooker
★ 5.1k🔥🔥 hooker is a Frida-based reverse engineering toolkit for Android. It offers a user-friendly CLI, universal scripts, auto hook generation, memory roaming to detect activities/services, one-click SOCKS5 proxy setup, Frida JustTrustMe, and BoringSSL u
Browse More Security MCP Servers
Explore all security servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up ToolTrust in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use ToolTrust?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.