SpecterAI

Attack Surface Intelligence Platform: recon tool powered by Claude AI.

Live Demo · Report a Bug · Usage

For authorized security testing only. Do not scan systems you don't own or don't have explicit written permission to test.

Screenshots

Dashboard	Live Scan Progress	Report View

The live demo runs on Render's free tier, so the first request after inactivity can take ~50s to wake up.

Table of contents

• What it does
• How it works
• Features
• Usage
  • CLI
  • Web dashboard
• Configuration
• Requirements
• Project structure
• License

What it does

Runs four recon modules in parallel against a target domain or IP, then sends the aggregated findings to Claude for AI-driven analysis and risk assessment. Outputs a structured Markdown report, or view it live in the web dashboard with real-time progress over SSE.

How it works

target ─┬─▶ DNS / WHOIS enumeration ─┐
        ├─▶ Port scan + banner grab  ├─▶ aggregator ─▶ Claude (risk analysis) ─▶ report
        ├─▶ HTTP header probe        │        (CLI: Markdown file · Web: live dashboard)
        └─▶ SSL/TLS inspection ──────┘

The four modules run concurrently, and their results are merged into one structured summary. That summary is sent to Claude, which returns a risk level, key findings, and next steps as structured JSON, which then gets rendered into the final report.

Features

• Four recon modules run in parallel: DNS/WHOIS, port scanning with banner grabbing, HTTP header analysis, SSL/TLS inspection
• AI-driven risk assessment via Claude: executive summary, key findings, and recommended next steps
• Live web dashboard with real-time scan progress over Server-Sent Events, plus scan history per browser session
• CLI mode for scripting or offline use (--no-ai skips the Claude call entirely)
• Quick or full scan modes: top 20 ports for a fast pass, top 1000 for deeper coverage
• Rate limited (5 scans per IP per minute on the web dashboard) to keep the demo usable for everyone
• Exportable reports: Markdown download or raw JSON via the dashboard

Usage

CLI

Install once, then run it as a regular command, no cloning required:

pipx install specter-ai   # or: pip install specter-ai
export ANTHROPIC_API_KEY=your_key_here

specter-ai --target example.com --mode quick
specter-ai --target example.com --mode full --output report.md
specter-ai --target example.com --no-ai   # skip AI analysis

Running from a source checkout instead of PyPI:

pip install -e .
export ANTHROPIC_API_KEY=your_key_here
specter-ai --target example.com --mode quick
# or: python -m specter_ai --target example.com --mode quick

Flags

Web dashboard

The web dashboard isn't published as a standalone package, so run it from a source checkout:

git clone https://github.com/hiratinspace/specter-ai.git
cd specter-ai
pip install -e ".[web]"
python3 web/app.py
# Open http://localhost:5000

Real-time scan progress via SSE, with a full report view and scan history on completion. Try it against a legal test target:

specter-ai --target scanme.nmap.org

Configuration

Requirements

• Python 3.10+
• ANTHROPIC_API_KEY environment variable (for AI analysis)

Core CLI dependencies (installed automatically via pip/pipx):

anthropic>=0.25.0
dnspython>=2.4.0
python-whois>=0.9.0
requests>=2.31.0

The web dashboard additionally needs the web extra:

flask>=3.0.0
gunicorn>=21.2.0

Project structure

src/specter_ai/       # Installable package (PyPI: specter-ai)
  cli.py              # CLI entrypoint (console script: specter-ai)
  modules/            # dns_enum, port_scan, http_probe, ssl_check
  core/               # aggregator, ai_analyst (Claude integration)
  report/             # Markdown report generator
web/app.py            # Flask web dashboard (run from a source checkout)
assets/screenshots/   # README screenshots

License

Apache License 2.0. See LICENSE for details.