A platform for discovering, fingerprinting, and security testing of exposed AI endpoints — MCP servers, Ollama, vLLM/LiteLLM proxies, LangServe chains, Gradio, and ComfyUI.
Works like Shodan for externally accessible AI services, identifying risks related to prompt leaks, CORS issues, unauthorized access, and agent‑tool exposure.

Features:
📍 Discovery of AI/ML endpoints via Shodan (more than 32 predefined queries).
📍 Fingerprinting using Nuclei templates and HTTP checks to detect frameworks, authentication, and models.
📍 Risk scoring (0–10) based on security parameters including TLS, CORS, and system prompt leakage.
📍 Protocol‑specific test scenarios (MCP tool abuse, Ollama model extraction, prompt injection)
📍 3D visualization of discovered services with Shodan‑like search.

Unlike general‑purpose scanners such as Nuclei or Shodan, AIMap focuses specifically on AI infrastructure, supporting protocol‑level tests and risk scoring but is less mature in terms of ecosystem development.

📎 Tool: https://github.com/BishopFox/aimap

💬 Discuss

AIMap — scanning and analysis of vulnerable AI services