Server facts
tooltrust-scanner MCP Server
Project
- Author
- AgentSafe-AI
- License
- Not specified
- Category
- Security & Reverse Engineering
Tech
- Runtime / language
- Not specified
- Transports
- stdio
- Remote available
- No
- Authentication
- Not specified
Activity
- GitHub stars
- 19
- Last commit
- Data as of
Tags
- mcp
- prompt injection
- scanner
- security
- supply chain
Overview
ToolTrust Scanner supports security review of MCP connections through live-server scans, configuration checks, and analysis of supplied tool definitions. Its live scan connects over stdio, inspects the exposed tools, and produces a risk report assigning each tool an ALLOW, REQUIRE_APPROVAL, or BLOCK policy. Users can also submit an MCP tools/list payload without using the live-server workflow. For Claude Code configurations, the scanner checks the current directory’s .mcp.json before falling back to ~/.claude.json, scans configured servers concurrently, and records startup failures while continuing with the others.
The checks cover injection attempts in tool descriptions, potential data exfiltration, permission overreach, and patterns associated with privilege escalation or arbitrary code execution. Other rules address compromised package releases, supply-chain vulnerabilities, misleading package names, tool shadowing, unsafe secret handling, and absent safeguards on risky tools. A rule-catalog tool exposes the checks and their identifiers. Directory lookup retrieves historical grades and a JSON scan report by server name, while the public browser interface lets users inspect findings and compare server grades before deciding what to trust or install.
tooltrust-scanner setup notes
- tooltrust-scanner records stdio only; confirm a network transport upstream before planning remote access.
- tooltrust-scanner has no recorded license; confirm reuse terms in its repository.