LYNX.
GitHub

See any company clearly.

Type a company. Fifteen agents research it across public sources, live.

Get started Star on GitHub
The Lynx dashboard for Anthropic: research agents on the left, an executive summary with legitimacy score, funding, valuation, SWOT and a timeline

One query, three ways to read it.

Every run fills a dashboard, a knowledge graph and a written documentary from the same cited facts.

Dashboard with an executive summary, scorecard, SWOT and timeline Knowledge graph centred on Anthropic, linking founders, investors, products and competitors A generated research documentary with wiki-linked entities and a research assistant chat panel

A card per category, filling in live as each agent lands. An executive summary sits on top: scorecard, SWOT and a merged timeline. A force-directed map of people, products, investors and competitors. Click any node for its facts, citations and connections. A long-form write-up with wiki-linked entities, plus a chat assistant that answers only from that run's data.

Built to show its work.

Every fact comes back with the source it came from, and every gap explains itself.

Source ladders

Agents walk a ranked list of sources, most authoritative first, and stop at the first rung that answers. Every rung reports what happened.

Key People, researching Anthropic
  1. Wikidata found 4 people
  2. Company site not needed
  3. Web search not needed

Fifteen agents, one job each

Each runs in isolation with its own timeout. A source going down costs you one card, never the run.

  • Overview
  • Profile
  • Stock
  • Financials
  • Funding
  • Products
  • Web presence
  • Key people
  • News
  • Community
  • Patents
  • Competitors
  • Legitimacy
  • Signals
  • Careers

Works without API keys

Every data source works out of the box. Add a free LLM key, like Groq, for richer analysis and a fully written documentary.

Measured, not promised.

Real runs against live sources, recorded by the benchmark harness in the repo. Every run finished all 17 passes.

18.2s for a complete run with no language model configured at all
Microsoft 124.4s
Duolingo 56.8s
Anthropic 43.1s
figma.com 31.6s
Anthropic, no LLM 18.2s
Windows 11, one API process, live network. Cache state and free-tier rate limits move these numbers. Full results

Data sources

Free, public sources first.

Metered services sit on the last rung of every ladder, so most runs never touch them.

Registries, filings and markets

  • SEC EDGAR
  • GLEIF LEI
  • RDAP
  • PatentsView
  • Yahoo Finance

Reference and archive

  • Wikipedia
  • Wikidata
  • Wayback Machine

Coverage and community

  • Google News
  • Hacker News
  • Reddit
  • X
  • Trustpilot

The company itself

  • Site crawl
  • Tech fingerprint
  • TLS handshake
  • DNS over HTTPS

Code and hiring

  • GitHub
  • Greenhouse
  • Lever
  • Ashby
  • SmartRecruiters

Search

  • DuckDuckGo
  • Firecrawl, optional

Running in a couple of minutes.

Lynx runs on your own machine. Your keys stay in your own .env and every run is saved to a local database.

  • Docker, or Node 20+ with pnpm and uv
  • Set CONTACT_EMAIL in .env (SEC asks for one)
  • Open localhost:3000 and type a company
git clone https://github.com/ronak-create/Lynx.git
cd Lynx
cp .env.example .env
docker compose up --build
git clone https://github.com/ronak-create/Lynx.git
cd Lynx
cp .env.example .env
cd apps/api && uv sync && cd ../..
pnpm install
pnpm dev

Point it at a company you're curious about.

Get started Star on GitHub