Aliens Eye is an AI-powered OSINT tool that hunts down social media accounts across hundreds of platforms using machine learning and heuristic detection.
The tool solves the problem of finding accounts associated with a given username across the internet. Rather than relying on simple HTTP status code checks, it combines a trained machine learning model with thirty structural signals—including HTTP status, DOM shape, keywords, and site-specific fingerprints—to detect whether a username exists on each platform. It scans over 840 platforms asynchronously and extracts profile metadata such as display names, bios, and avatars from detected accounts. The tool can correlate profiles that appear to belong to the same person by comparing avatar hashes, bio text, shared links, and names. It supports recursive expansion to follow usernames mentioned in bios and re-scan them, domain availability checking, watch mode for periodic re-scanning with webhook notifications, and resumable scans that checkpoint progress.
The tool suits security researchers, investigators, and penetration testers who need to map a person's online presence across many platforms at once. It is particularly valuable for cases where naive status-code detection produces too many false positives or misses accounts on JavaScript-heavy sites. The project includes a self-check command that reports precision, recall, F1 score, and false positive rate per site, and supports reproducible evaluation by recording and replaying frozen response corpora. It can be compared against Sherlock, Maigret, and WhatsMyName through built-in evaluation commands. The tool offers modern terminal UI with live progress and result tables, optional interactive browser mode, and an MCP server for exposing scanning to LLM agents. It supports proxies, Tor, site filtering, and output in multiple formats including JSON, CSV, HTML, Markdown, PDF, and graph formats.
The project maintains a continuous integration pipeline. The codebase includes ablation testing and active learning capabilities, allowing users to retrain the detector and hand-label uncertain results. A self-check mechanism provides calibrated performance metrics across individual sites.
