AI-Surface

•
145
•

Identifies and governs AI attack surfaces within application code during development and integration stages.

AI-Surface is a free and open-source static analysis tool designed to map and manage the AI attack surface in your codebase. It detects and flags various AI components such as LLM calls, agents, Model Context Protocol (MCP) servers, RAG/vector stores, model gateways, and self-hosted runtimes, along with any HTTP APIs that expose them. Running locally and offline, it helps engineering and security teams identify AI-related risks before code merges, generate an AI-BOM, and ensure compliance with standards like OWASP LLM Top 10, EU AI Act, NIST AI RMF, and ISO 42001.

主な機能

01131 GitHub stars
02Operates as a local, offline static analysis tool with no telemetry, ensuring source code privacy.
03Identifies AI attack surfaces including LLM calls, agents, MCP servers, RAG/vector stores, and model gateways.
04Integrates seamlessly into CI/CD pipelines via GitHub Action for PR-time gating on new high-risk findings.
05Generates an AI-BOM and provides governance evidence mapped to OWASP LLM Top 10, EU AI Act, NIST AI RMF, and ISO 42001.
06Offers an interactive local UI for visualizing detected AI surfaces, risk indicators, and governance badges.

ユースケース

01Reviewing AI-related pull requests before they merge to gate new high-risk findings.
02Building an AI-BOM or AI-governance inventory from source code.
03Adding agents, MCP servers, RAG, model gateways, or LLM calls to an application.