AI Scanner
Run an agent-driven pass over your repo to find LLM SDK usage, leaked API tokens, and hardcoded secrets before release.
Overview
AI Scanner MCP is a Ship-phase MCP server that scans repositories for LLM SDK usage, exposed API tokens, and hardcoded secrets.
What is this MCP server?
- Scans codebases for LLM and AI SDK adoption patterns via ai-scanner-mcp
- Flags exposed API tokens and hardcoded secrets suitable for agent-led audits
- stdio npm package ai-scanner-mcp version 1.0.3; server version 1.0.6
- GitHub source at Aakashbhardwaj27/ai-scanner-mcp
- Developer security MCP—not a production monitoring dashboard
- Server version 1.0.6 on MCP registry
- npm identifier ai-scanner-mcp at package version 1.0.3
- stdio transport; GitHub repository ai-scanner-mcp
Community signal: 1 GitHub stars.
What problem does it solve?
Agent-heavy repos accumulate AI SDK imports and .env-style leaks that are easy to miss in a solo pre-ship review.
Who is it for?
Solo builders shipping LLM-powered apps who want MCP-native secret and SDK discovery in the same session as code fixes.
Skip if: Organizations that already run enterprise SAST with policy gates only, or repos with no AI dependencies to audit.
What do I get? / Deliverables
Your agent can run structured scans and surface secret and AI-usage risks before you tag a release or open-source the tree.
- Agent-triggered scan results for AI SDK usage
- Reported exposed tokens and hardcoded secret locations
- Actionable file-level leads for remediation before ship
Recommended MCP Servers
Journey fit
How it compares
Focused codebase secret and AI-SDK scanner MCP, not cloud runtime monitoring or generic pentest automation.
Common Questions / FAQ
Who is AI Scanner MCP for?
Developers using Claude Code or similar agents who need quick LLM-footprint and secret-leak scans on their own repositories.
When should I use AI Scanner MCP?
Use it in Ship security checks before releases, public repo pushes, or when onboarding a new contractor to an AI codebase.
How do I add AI Scanner MCP to my agent?
Install ai-scanner-mcp from npm, point your MCP client at the stdio server, and invoke scan tools against your project root from the agent.