ARGUS - Research Orchestrator

Not a search engine, not a scraper, not a wrapper. A research orchestration skill and CLI that routes any AI agent through existing open-source tools using a 5-wave methodology with cross-verification.

Philosophy: do not reinvent the wheel. Compose existing tools (yt-dlp, Firecrawl, Crawl4AI, TEx, Telerecon) into one coherent research workflow.

Quick start

ARGUS: <query>
ARGUS: compare Notion vs Obsidian
ARGUS: platform AI agents on Twitter
ARGUS: timebox EU AI Act 2024-2026
ARGUS: region AI market in Africa
ARGUS: narrow pricing model of Stripe
ARGUS: deep quantum computing startups in Europe
ARGUS: quick who is the CEO of X

Local CLI (standard library only, no install required):

python argus_search.py plan "ARGUS: compare Notion vs Obsidian"
python argus_search.py report "ARGUS: deep AI in Africa" --output ./output
python argus_search.py preview durov --limit 5
python argus_search.py verify ./output/argus-report-<slug>-<date>.md   # mandatory gate before delivery
python argus_search.py modes

Install as a package with MCP server (for agents):

pip install .              # CLI: argus
pip install .[mcp]         # MCP server: argus-mcp

MCP deployment: run the MCP server on stdio or localhost only. Never bind it to a network socket without authentication, TLS, and a path allowlist, because generate_report and verify_report perform filesystem I/O.

Contents

Golden rule

No scraping, no forums or Telegram deep search, no dark web, no restricted sources - without asking the user first and getting an explicit [Y]. Silence means SKIP, always.

Status

• Phase 1 (build) done. Phase 2 (verify) done. Phase 3 (publish prep) done.
• Verified 2026-10-07: 7/7 modes, 7/7 routing entries, 4/4 consents, 5/5 verification checks.
• Full Telegram collection requires the user's own TELEGRAM_API_ID and TELEGRAM_API_HASH.