Skip to main content
 █████╗ ██████╗  ██████╗ ██╗███████╗
██╔══██╗██╔══██╗██╔════╝ ██║██╔════╝
███████║██████╔╝██║  ███╗██║███████╗
██╔══██║██╔══██╗██║   ██║██║╚════██║
██║  ██║██║  ██║╚██████╔╝██║███████║
╚═╝  ╚═╝╚═╝  ╚═╝ ╚═════╝ ╚═╝╚══════╝

  the all-seeing OSINT collector   ·   v0.9.1

Argis

PyPI version Downloads License Tests

The all-seeing OSINT scanner. Map a username across 500+ platforms with media enrichment, identity correlation, breach checks, impersonation detection, and footprint drift tracking — in one async CLI.

Named after Argus Panoptes, the hundred-eyed giant of Greek myth: every platform, watched at once.

pip install argis
argis scan johndoe

Table of Contents


Demo

Argis scan demo

argis scan johndoe — 509 platforms, async, streamed to the terminal as results land.

The GIF above is a stylized mockup for illustration. For real footage of your own build, record one with asciinema, VHS, or Terminalizer and drop it in at assets/demo.gif.


Why Argis

Sherlock and Maigret answer does this handle exist here? Argis answers everything after that: it correlates identities, flags impersonators, tracks drift over time, self-heals its own site rules, exports to HTML/PDF/Neo4j/XMind, and layers in recon, breach, and media tooling — one pipeline instead of five separate tools.

Feature Argis Sherlock Maigret
Platforms scanned 509 ~400 ~350
Identity correlation (link) ✅ ❌ ❌
Impersonation detection (guard) ✅ ❌ ❌
Identity drift tracking (echo) ✅ ❌ ❌
Self-healing site rules (doctor) ✅ ❌ ❌
HTML / PDF / XMind / Neo4j exports ✅ ❌ ❌
Media enrichment + avatar classifier ✅ ❌ ❌
Recon: ports, DNS, WHOIS ✅ ❌ ❌
OCR from screenshots ✅ ❌ ❌
Face detection + reverse search ✅ ❌ ❌
Breach checking ✅ ❌ ❌
Change monitoring ✅ ❌ ❌
LLM risk analysis ✅ ❌ ❌
MCP server (AI integration) ✅ ❌ ❌
Web browser UI ✅ ❌ ❌
Docker support ✅ ✅ ✅

Use Cases

Personal security audit — argis me you shows everywhere your handle exists, checks for breached emails and impersonators, and scores your exposure with a ranked shrink plan.

Continuous monitoring — argis monitor johndoe --interval 86400 --webhook <url> watches an account daily and pushes an alert the moment something changes.

Brand & handle protection — argis guard yourbrand generates the full homoglyph / typo-squat space around your handle and flags lookalike accounts before they gain traction.

Authorized security investigations — scan a handle in scope for your engagement, pipe results into Neo4j, cross-reference via the MCP server, and export a PDF for case documentation.

OSINT pivoting — resolve a handle's identity cluster, outbound links, and geo-hints to map how a public account connects to the rest of its footprint.

Infrastructure recon — argis recon example.com runs port scanning, service detection, DNS resolution, WHOIS, and geolocation from one CLI, for hosts you own or are authorized to assess.


Quick Start

pip install argis
argis --version                           # confirm the install
argis scan johndoe                        # surface every account
argis scan johndoe --dossier report.html  # build the full HTML dossier
argis scan johndoe -P report.pdf          # PDF report
argis me johndoe                          # full self-assessment
argis investigate johndoe --html report.html # run 50-agent deep investigation
argis echo johndoe                        # track identity drift over time
argis web                                 # launch the browser UI

Installation

pip install argis                    # base install
pip install "argis[web]"             # + web UI (uvicorn)
pip install "argis[intel]"           # + avatar matching for link/guard
pip install "argis[screenshots]"     # + OCR and screenshots
pip install "argis[vision]"          # + face detection
pip install "argis[insightface]"     # + offline DeepFace matching
pip install "argis[render]"          # + headless browser for JS-gated profiles
pip install "argis[mcp]"             # + MCP server
pip install "argis[pdf]"             # + PDF generation
pip install "argis[dev]"             # + test suite
pip install "argis[all]"             # everything

Requires Python 3.10+. Supports Windows, Linux, and macOS.


Features

🔍 509 platforms Social, coding, gaming, creative, professional, and more
⚡ Async engine Concurrent scanning with HTTP/2 support and retry-with-backoff
🖼️ Media enrichment Captures real profile photos (GitHub & Instagram first-party APIs) via a classifier that separates true avatars from logos, favicons, and generic Open Graph art
📊 HTML dossier Risk banner, distribution, identity, correlations, captured media, filterable account list — plus PDF, TXT, CSV, Markdown, XMind, GraphML, Neo4j exports
🔁 Echo identity drift Detects coordinated rebrands, avatar migrations, contact pivots, account expansion/retreat across saved scans
🖥️ Recon Nmap-style port scan, service/version detection, OS detection, UDP scan, traceroute, host discovery
🌐 DNS & WHOIS Resolve records, look up domain ownership
🌎 Geolocation IP geolocation via ipgeolocation.io
🕒 History & diff Track changes to a footprint over time
🛰️ Change monitoring Continuously watch usernames and report changes
↔️ Side-by-side comparison Compare two usernames
🎬 Wayback Machine Historical snapshots of profiles
📁 Multiple outputs JSON, CSV, HTML, Markdown, TXT, NDJSON, XMind, GraphML, Neo4j, PDF, webhooks (Slack / Discord)
🧠 AI analysis LLM-powered risk assessment via OpenAI or Anthropic
🕵️ Deep Investigation Orchestrates 50 specialized AI agents across 5 squads (Identity, Social, Professional, Deep Web, Specialists)
🤖 OCR Extract usernames from screenshots, auto-scan found handles
📷 Face detection Reverse-search via multiple engines (Google, TinEye, Bing, Yandex, SauceNAO, IQDB, ImgOps); offline DeepFace lookalike matching
🧹 Self-healing Auto-verifies site rules and flags silent rot
🔗 Identity correlation Clusters accounts into real identities vs. impersonators
🛡️ Impersonation guard Hunts lookalike handles wearing your face
🔒 Breach checker Checks whether emails were exposed in known breaches
💬 Web mentions Searches pastes, code, and dorks for a handle
👤 Unified threat report argis me consolidates your entire footprint
🔌 MCP server Connects Argis to any MCP-compatible AI assistant
🌐 Web UI Browser mode with live streaming results, category-grouped cards, inline dossier
📸 Screenshots Captures profile pages via Playwright, renders as ANSI art in-terminal
🖥️ Desktop notifications Push notifications on scan completion

Intelligence

Every username scanner answers one question: does this handle exist here? Argis Intelligence answers what comes after.

Command What it answers
investigate Deep multi-agent investigation utilizing 50 specialized AI agents to analyze a target's footprint.
doctor Are my detection rules still correct, or have they silently rotted?
link Of everywhere this handle exists, which accounts are the same person — and which are impostors?
guard Is anyone impersonating me on a lookalike handle right now?
echo Did this identity rebrand, migrate avatars, or retreat across platforms together?

These tools are built for defensive / self-OSINT: verifying your own data, disambiguating accounts that already share your public handle, and surfacing people impersonating you — see Responsible Use for the intended scope.

investigate — deep multi-agent intelligence

Runs a comprehensive analysis on a target using 50 specialized AI agents organized into 5 squads (Core Identity, Social Intel, Professional, Deep Web, and Specialists). Generates an advanced HTML dossier or Markdown report aggregating all AI agent findings.

argis investigate johndoe --html report.html

doctor — self-healing site database

sites.json rots silently: a platform tweaks its markup or 404 behavior, a rule breaks, and nobody notices for months. doctor re-runs every rule against a known-real and known-fake username and flags what's broken.

argis doctor                                # health-check every rule
argis doctor --only GitHub,Reddit,Steam     # spot-check a few
argis doctor --report health.md --json health.json --exit-code

Ships with a weekly GitHub Action. Also flags duplicate rule names in sites.json.

Runs a scan, pulls each found profile's avatar, display name, bio, outbound links, and emails, then clusters accounts into identity groups. The largest cluster is you; anything sharing the handle but scoring below threshold is flagged as a namesake or impersonator.

argis link johndoe
argis link johndoe --threshold 0.7 --category social,media
argis link johndoe --no-avatar

Scoring blends avatar perceptual hash (dHash), name/bio similarity, shared links, and shared emails.

guard — impersonation early warning

Nobody impersonates you with your exact handle: they register john_doe, j0hndoe, johndoe_official, or the homoglyph jоhndoe (Cyrillic o). guard generates the confusable space around your handle, scans every variant, and scores each hit against your real profile.

argis guard johndoe --list                   # preview the variant space
argis guard johndoe --reference https://github.com/johndoe
argis guard johndoe --threshold 0.65 --category social

echo — coordinated identity drift

A normal diff compares two scans. echo analyzes the full saved history and groups changes that happened in the same window: rebrands, avatar migrations, contact pivots, multi-platform expansion or retreat.

argis scan johndoe          # baseline, then scan again later
argis echo johndoe
argis echo johndoe --window 24 --min-confidence 70
argis echo johndoe --json -o johndoe-echo.json

Reports an identity stability score, identity epochs, event confidence, affected platforms, and full before/after evidence in JSON.


Media Pipeline

Argis captures profile photos and classifies each image so the dossier shows real avatars, not platform chrome.

argis scan johndoe --dossier report.html     # dossier with captured media
argis media-review johndoe                   # interactive confidence dashboard (opens browser)
argis media-apply johndoe-media-review.json  # save your accept/reject decisions
argis scan johndoe --dossier final.html      # dossier now uses only approved media
argis media-clear                            # reset all decisions to automatic

Each image is classified as PROFILE_AVATAR, PROFILE_BANNER, PLATFORM_LOGO, GENERIC_THUMBNAIL, DEFAULT_AVATAR, UNKNOWN_MEDIA, or REJECTED. Only validated profile avatars count toward stats, correlation, and risk.


Commands

Discovery

Command Description
scan Search a username across 509 platforms (add --dossier for the HTML report)
scan-image OCR a screenshot for usernames/URLs
scan-face Detect faces and reverse-search across multiple engines
setup-celebrity-db Download celebrity face data for offline DeepFace matching

Intelligence

Command Description
investigate Deep multi-agent investigation across 50 specialized AI agents (5 squads)
me Unified self-assessment: scan + breach + mentions + geo + impersonation
breach Check emails for known breaches
mentions Search pastes, code, and dorks for a handle
locate Infer geographic region from profile signals
link Cluster accounts into real identities vs. impostors
guard Hunt lookalike handles impersonating you
doctor Health-check every site rule and flag rot

Reconnaissance

Command Description
recon Port scan, service detection, OS fingerprinting, DNS, WHOIS, geo
discover Sweep a subnet to find live hosts
domain DNS resolution, WHOIS, port scan
myip Show public IP + geolocation

Tracking

Command Description
history Show past scan timestamps and found-profile counts
clear-history Delete saved scan history for a username
monitor Continuously watch a username for changes
echo Detect coordinated identity drift across saved scans

Analysis

Command Description
exposure Privacy risk score (0–100), grade (A–F), ranked shrink plan
timeline Chronological timeline of when accounts were created
graph Interactive pivot graph from a seed handle
compare Compare two usernames side by side
wayback Wayback Machine snapshots of profiles
media-review Interactive media confidence dashboard
media-apply Save media approvals for future dossiers
media-clear Reset saved media decisions

Utilities

Command Description
web Launch the local Argis web UI (browser mode)
mcp Run Argis as an MCP server
search Full-text search across scan history
stats Aggregate scan statistics across tracked users
categories List all platform categories with counts
import-sites Import Sherlock/Maigret site databases
Full flag reference (click to expand)

scan

argis scan johndoe                            # basic scan
argis scan johndoe --category coding,social   # filter by category
argis scan johndoe --site GitHub              # just one platform
argis scan johndoe --exclude Facebook,Twitter # skip specific platforms
argis scan johndoe --status FOUND             # show found only
argis scan johndoe --min-confidence 60        # only high-confidence hits
argis scan --file usernames.txt --export csv  # batch scan
argis scan johndoe --diff                     # compare vs last scan
argis scan johndoe --emails                   # extract emails
argis scan johndoe --notify                   # desktop notification
argis scan johndoe --proxy socks5://127.0.0.1:9050
argis scan johndoe --tor --timeout 15 --concurrency 10
argis scan johndoe --http2                    # HTTP/2 multiplexing
argis scan johndoe --retry --no-retry         # control retry behaviour
argis scan johndoe --json-stream              # JSON lines output
argis scan johndoe --screenshots              # capture profile page screenshots
argis scan johndoe --screenshots --show       # render in terminal
argis scan johndoe --list                     # list platforms to scan (dry)
argis scan johndoe --dossier report.html      # full HTML dossier
argis scan johndoe -P report.pdf              # PDF report
argis scan johndoe -T report.txt              # TXT report
argis scan johndoe -C report.csv              # CSV report
argis scan johndoe -H report.html             # HTML report
argis scan johndoe -M report.md               # Markdown report
argis scan johndoe -X report.xmind            # XMind mindmap
argis scan johndoe -G graph.graphml           # GraphML export
argis scan johndoe --neo4j import.cypher      # Neo4j import script
argis scan johndoe -J ndjson                  # NDJSON export
argis scan johndoe --ai                       # AI-powered risk analysis
argis scan johndoe --ai-model claude-3-opus   # custom AI model
argis scan johndoe --webhook https://discord.com/api/webhooks/...

scan-image / scan-face / setup-celebrity-db

argis scan-image screenshot.png               # OCR for usernames/URLs
argis scan-image screenshot.png --scan        # auto-scan extracted handles

argis scan-face photo.jpg                     # detect faces
argis scan-face photo.jpg --search             # open reverse search in browser
argis scan-face photo.jpg --search --engine tineye
argis scan-face photo.jpg --identify           # auto-identify + scan
argis scan-face photo.jpg --identify --offline # DeepFace offline only
argis scan-face photo.jpg --crop               # save face crops

argis setup-celebrity-db                       # download reference images
argis setup-celebrity-db --force                # redownload if cached

me / echo

argis me johndoe                              # full assessment
argis me johndoe --skip-impersonation         # skip lookalike scan
argis me johndoe --max-variants 30

argis echo johndoe
argis echo johndoe --window 24 --min-confidence 70
argis echo johndoe --json -o johndoe-echo.json

recon

argis recon example.com                       # basic scan
argis recon example.com -sv -os -df -tr       # version + OS + scripts + traceroute
argis recon example.com -ax -pt '*'           # aggressive, all ports
argis recon example.com -gl                   # geolocation
argis recon -ag 192.168.1.0/24                # ping sweep
argis recon -ud example.com                   # UDP scan
argis recon -ox scan.xml -on scan.txt example.com

monitor / web / mcp

argis monitor johndoe --interval 30           # check every 30 seconds
argis monitor --file users.txt --webhook https://hooks.slack.com/...

argis web                                     # launch on http://127.0.0.1:8000
argis web --host 0.0.0.0 --port 8080

argis mcp                                     # stdio (Claude Desktop / Claude Code)
argis mcp --transport sse --port 8080         # SSE (web clients)

Documentation

Full docs at mohilisop.github.io/argis | Website: argis.dev


Responsible Use

Argis is built for auditing your own exposure, defending your own handle or brand, and authorized security work — not for tracking private individuals without their knowledge or consent.

  • Only run scan, link, guard, echo, monitor, or scan-face --identify against handles, faces, or domains you own or are explicitly authorized to assess.
  • Depending on your jurisdiction, using these features against someone else without consent may violate anti-stalking/harassment law, computer-misuse statutes (e.g. the U.S. CFAA, UK Computer Misuse Act), platform terms of service, or data-protection law (e.g. GDPR/CCPA) — that risk sits with the operator, not with Argis.
  • link and guard are identity-disambiguation tools (which accounts are you vs. an impostor), not deanonymization tools for figuring out who an anonymous stranger is.
  • Respect target sites' robots.txt and rate limits; Argis's retry/backoff is there to be a good citizen of the platforms it queries, not to defeat their protections.
  • The maintainer does not support, and takes no responsibility for, use of Argis for stalking, harassment, doxxing, or unauthorized surveillance.

If you maintain a fork or downstream tool built on Argis, please carry this section forward.


Contributing

Contributions are welcome:

  1. Fork the repo and create a feature branch (git checkout -b feature/my-change).
  2. Run argis doctor and the test suite before opening a PR — pip install "argis[dev]" && pytest.
  3. Keep new site rules in sites.json covered by a doctor health check.
  4. Open a PR describing what changed and why.

Bug reports and platform-rule fixes (a site rule silently broke) are especially appreciated — argis doctor --report health.md is the fastest way to spot those before you file.


Security

Argis is designed for defensive / self-OSINT use. If you discover a security vulnerability, please do not open a public issue — see SECURITY.md and security.txt for responsible-disclosure instructions and contact details.


Star History

Star History Chart


License

MIT

Metadata

Release files for argis 0.9.2

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for argis 0.9.2
File Size Uploaded Provenance
argis-0.9.2.tar.gz 1.3 MB Without provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for argis 0.9.2
File Interpreter ABI Platform
argis-0.9.2-py3-none-any.whl Python 3 none any Details

Total release size: 1.5 MB

Release files / argis-0.9.2.tar.gz

Download URL argis-0.9.2.tar.gz
Size 1.3 MB
Tags Source
SHA-256 checksum
How to use checksums
7c1c8181264652f8330561a5876660637482777f402fe74a370a9fa88a5519fe
BLAKE2b-256 checksum
How to use checksums
d14696856263ba5f7545fea849988801a861f8386a880dc2f35b99d347a13d3f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.13.12

Release files / argis-0.9.2-py3-none-any.whl

Download URL argis-0.9.2-py3-none-any.whl
Size 275.3 kB
Tags Python 3
SHA-256 checksum
How to use checksums
13918b3d800244a53812f8304517475d41d576662728db0b489629ee663a2407
BLAKE2b-256 checksum
How to use checksums
0e73e990d01b9e4fefeb053922d916b1ae4c6d105a9eb676d30a82a1b5e4e82c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.13.12

Security for argis 0.9.2

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page