TraceHunt MCP Server

TraceHunt is a privacy-respecting, self-hostable fork of the well-known Sherlock project (MIT). It keeps Sherlock's battle-tested detection engine and site database, and adds reporting, scoring, configuration and a no-phone-home default. See NOTICE.md for full attribution and the list of changes.

For authorized security research and OSINT only. Check the laws in your jurisdiction and only investigate accounts you are permitted to.

繁體中文摘要

TraceHunt 是一套 OSINT 使用者名稱偵查工具：輸入一個帳號名稱，它會在 480+ 個 社群/平台上查詢該名稱是否被註冊，協助資安研究與數位足跡盤點。

本專案是知名開源工具 Sherlock 的客製化分支（MIT 授權），在保留原本偵測引擎與 網站資料庫的基礎上，新增了以下功能：

• HTML 報告：一鍵產生美觀、可離線開啟的單檔報告（--html report.html）。

• 數位足跡評分：統計命中數並換算 0–100 分的足跡分數（--summary）。

• YAML 設定檔：把常用參數寫進 tracehunt.yaml，免得每次重打（--config）。

• 隱私優先：原版啟動時會自動連線檢查更新，本版改為 預設不連外，只有加上 --check-update 才會檢查。

授權與致謝請見 NOTICE.md；原始著作權保留於 LICENSE。

Related MCP server: Maigret MCP Server

Features

Install

git clone https://github.com/Hayatelin/tracehunt.git
cd tracehunt
pip install -r requirements.txt
# optional: install as a CLI command
pip install .

Requires Python 3.9+.

Quickstart

# Basic: hunt one username, print results to the terminal
python -m tracehunt johndoe

# Generate a styled HTML report + a footprint summary
python -m tracehunt johndoe --html johndoe.html --summary

# Limit to specific sites and export CSV
python -m tracehunt johndoe --site GitHub --site Reddit --csv

# Use a config file for your default options
python -m tracehunt johndoe --config tracehunt.yaml

If you installed it as a command, replace python -m tracehunt with tracehunt.

🤖 Use it from your AI agent (MCP + skill)

TraceHunt ships with an MCP server and an Agent Skill, so Claude Code, Cursor, Codex or Gemini CLI can run username recon for you on demand.

pip install "mcp[cli]" -r requirements.txt
python mcp/tracehunt_mcp.py        # exposes hunt_username() + footprint_score()

Register it with your agent (see mcp/README.md), then just ask: "use tracehunt to check the username octocat". Prefer the no-MCP route? The skill/SKILL.md teaches an agent to drive the CLI directly. There's also a plain Python API: from tracehunt.api import hunt.

Architecture

flowchart LR
    A[CLI args + tracehunt.yaml] --> B[core.run_search]
    B --> C[(site database<br/>data.json)]
    B --> D[Async requests<br/>requests-futures]
    D --> E[QueryResult per site]
    E --> F{Outputs}
    F --> G[Terminal]
    F --> H[CSV / XLSX]
    F --> I[report.py<br/>HTML + footprint score]

What's different from Sherlock

The detection engine and site list come from Sherlock (MIT). The added value:

• tracehunt/report.py — HTML report + footprint scoring (new, stdlib only)

• tracehunt/config.py — YAML config loader (new)

• core.py — new --html / --summary / --config / --check-update flags; the automatic update check is now opt-in, and the bundled offline DB is the default.

Full details and attribution: NOTICE.md.

Tests

pip install pytest PyYAML
pytest

License

MIT — see LICENSE. Original work © 2019 Sherlock Project; TraceHunt modifications © 2026 VictorLin.