Scan your local codebase for security vulnerabilities, hardcoded secrets, and insecure coding patterns — all from your AI assistant via MCP (Model Context Protocol).
- AWS Access Keys & Secret Keys
- GitHub tokens (personal, OAuth, app)
- Stripe API keys (live/test)
- Slack tokens & webhooks
- Google Cloud / Firebase credentials
- Database connection strings
- JWT tokens & private keys (RSA, DSA, EC)
- npm auth tokens, Telegram bot tokens, SendGrid API keys
- Generic API keys & password assignments
Automatically detects and parses:
package.json(npm/yarn/pnpm)requirements.txt,Pipfile,pyproject.toml(Python)go.mod(Go)Cargo.toml(Rust)pom.xml,build.gradle(Java)
Checks against a built-in database of 45+ CVEs across JavaScript, Python, Java, Go, and Rust ecosystems.
- SQL Injection: String concatenation in queries, raw SQL builders
- XSS: innerHTML, dangerouslySetInnerHTML, v-html
- Command Injection: os.system, subprocess shell=True, eval/exec, child_process.exec
- Path Traversal: Unsanitized file paths
- Insecure Deserialization: pickle, yaml.load, marshal
- Configuration Issues: Debug mode, CORS wildcard, hardcoded JWT secrets
- Information Leakage: Stack trace exposure, directory listing
| Tool | Description |
|---|---|
scan_secrets |
Scan for hardcoded API keys, tokens, and passwords |
scan_dependencies |
Check dependencies against known vulnerability database |
scan_code_patterns |
Detect SQLi, XSS, command injection, and other patterns |
scan_file |
Comprehensive scan of a single file (secrets + code patterns) |
scan_directory |
Full project audit (secrets + dependencies + code patterns) |
- Python 3.11+
pip install mcp pydantic
git clone https://github.com/214070779/code-scanner-mcp.git
cd code-scanner-mcp
pip install mcp pydantic
npx @modelcontextprotocol/inspector python3 server.pyAdd to your MCP settings:
{
"mcpServers": {
"code-scanner": {
"command": "python3",
"args": ["/path/to/code-scanner-mcp/server.py"]
}
}
}"Scan my project for security issues"
→ AI calls scan_directory(path="./my-project")
"Check this file for secrets before committing"
→ AI calls scan_file(path="./src/config.ts")
"Are there any vulnerable npm packages?"
→ AI calls scan_dependencies(path=".")
# Clone and install
git clone https://github.com/214070779/code-scanner-mcp.git
cd code-scanner-mcp
pip install mcp pydantic
# Run tests
python3 -c "from server import mcp; print('OK:', list(mcp._tool_manager._tools.keys()))"
# Run with inspector
npx @modelcontextprotocol/inspector python3 server.pyMIT