# MEC Co (mec-co.jp) — Comprehensive Intelligence Gathering

## Goal
Exhaustively collect everything that has been said or is known about MEC Company (mec-co.jp) — a Japanese specialty chemicals company. Not much exists publicly, so the bar is: really everything, with special tooling for sources we'll likely need again.

## Scope
- **Company info**: Products, financials, management, history, patents, customers
- **Japanese sources**: IR pages, press releases, Nikkei, Kabutan, Yahoo Finance Japan, TDB (Teikoku Databank)
- **English sources**: Any coverage — trade publications, chemical industry reports, supply chain mentions
- **Patent databases**: JPO (Japan Patent Office), Google Patents — MEC Co chemical formulations
- **Trade show / conference mentions**: SEMICON, JPCA Show, etc.
- **LinkedIn / company profiles**: Key employees, org structure
- **Regulatory filings**: EDINET (Japanese SEC equivalent), annual reports (有価証券報告書)

## Special Tooling Needed (reusable for future Japanese company research)
1. **EDINET scraper**: Japanese company filings (equivalent to EDGAR) — `disclosure.edinet-fsa.go.jp`
2. **Kabutan profile extractor**: `kabutan.jp/stock/?code=XXXX` — financials, news, analyst coverage
3. **Yahoo Finance Japan**: `finance.yahoo.co.jp` — stock data, news, message boards
4. **Google Patents JP**: Patent search filtered by applicant
5. **Japanese IR page crawler**: Company IR pages often have PDF annual reports, presentations
6. **Translation layer**: Japanese → English for all extracted content

## Stages
1. **discover_ir** — Crawl mec-co.jp IR pages, download all PDFs, presentations
2. **discover_filings** — EDINET search for MEC Co filings
3. **discover_patents** — Google Patents search for MEC Co applicant
4. **discover_news** — Japanese + English news search
5. **discover_trade** — Industry publication search (chemical, PCB, semiconductor)
6. **fetch** — Download all discovered content (brain/ for web, direct for PDFs)
7. **translate** — LLM translate Japanese content to English
8. **extract** — Structured extraction: products, financials, customers, technology
9. **compile** — Company profile document with all findings

## Storage
```
data/companies/mecco/
├── ir/                    # IR page content, PDFs
│   ├── annual_reports/
│   ├── presentations/
│   └── press_releases/
├── filings/               # EDINET filings
├── patents/               # Patent documents
├── news/                  # News articles (JP + EN)
├── trade/                 # Trade publication mentions
├── extracted/             # Structured extractions
├── profile.md             # Compiled company profile
├── sources.yaml           # All source URLs with fetch dates
└── index.jsonl
```

## Reusable Components (build as general tools)
- **EDINET client**: Query + download filings by company code → `tools/edinet/`
- **Kabutan scraper**: Company profile + financials → `tools/kabutan/`
- **JP IR crawler**: Generic Japanese company IR page crawler → via `browser/`
- **JP→EN translation pipeline**: Batch translate with LLM, cache results

## Key Considerations
- **MEC Co stock code**: Need to identify their TSE code (likely 4971)
- **mec-co.jp may require JS rendering**: Use playwright-servers for crawling
- **PDF extraction**: Annual reports are PDF — use brain/ PDF extraction
- **Rate limiting on Japanese sites**: Be respectful, 1 req/5s on kabutan/yahoo
- **Archival**: Wayback Machine for historical IR pages that may have been removed

## Pacing
- Japanese sites: 12/hr (conservative, respectful)
- EDINET API: 30/hr
- LLM translation: 30/hr (longer texts)
- Web search: 20/hr