Custom links

Any URL you paste here gets scraped on each sync. Claude reads the page + extracts up to 20 recent news items as raw signals.

Where to add

• Per company — in the company page or company detail modal
• Org-wide — Settings → Custom links. Items match to companies by name mention

How it works

1. Fetch HTML (timeout 15 s, 50 KB cap)
2. Strip nav/footer/scripts
3. Send to Claude Haiku 4.5 with a fixed extraction prompt
4. Return: [{title, date, snippet, url}]
5. Cache by content hash — if the page didn't change since last sync, skip the Claude call entirely
6. Items get fed into the normal classify pipeline

Signal-type tags

Each link can be tagged with specific signal types. If you tag a link as funding, items classified as hiring / partnership / etc are dropped — the link is a funding-only source.

Leave tags empty = accept all types from that link.

Cost

Cold scrape ≈ $0.012 (~10 K input tokens to Haiku). Subsequent unchanged-page syncs = free thanks to the cache.

When to use

• Industry news sites that mention many target companies (org-wide)
• A specific customer's blog / press page (per-company)
• Aggregator feeds that don't have RSS