Free OSINT Tools That Actually Work: A Job-by-Job Map (Free Tiers, Real Ceilings)

The honest answer about free OSINT tools: the tools are rarely the hard part. Collection at scale, dedupe, provenance, and alerting are. This list is grouped by job, not by brand, because "which tool" is the wrong question — "which job am I paying for" is the right one.

Collection (free tiers that are real)

• t.me/s/<channel> web preview — public Telegram channels as plain HTML, zero credentials. The best free collection layer most people ignore. My collector design for it: https://dev.to/yuhehe/telegram-scraping-in-python-without-an-api-key-the-public-preview-collector-that-survives-12ef
• Shodan (free tier) — internet-wide asset census. The free 327K-IP query pool is a census sample, not the census; design queries knowing the ceiling.
• Certificate Transparency logs (crt.sh) — subdomain discovery from cert issuance. Free, complete for what it is, and quietly one of the highest signal-to-noise discovery sources.
• Wayback Machine API — historical snapshots: "what did this page claim in March" is a free, underused question.
• GitHub/Gist code search — leaks, config drift, internal hostname sprawl. Free within rate limits; your own regex layers beat default secret patterns.

Extraction & parsing

• Readability-style extractors (trafilatura, readability-lxml) — turn HTML into text you can diff. Text is what you alert on.
• yt-dlp for video provenance metadata, exiftool for media forensics on anything user-uploaded.

Analysis (the layer that pays)

This is where free tools stop helping, because the work is yours:

• Baselines before alerts. Subscriber counts, posting cadence, wording frequencies — measured over weeks, not scraped in one shot.
• Change detection, not new-arrival detection. Alert on the sentence that changed, not the document that arrived.
• Corroboration rules. Two independent sources = signal. One source, five mirrors = noise with a fan club.

What the paid tools are actually selling

Not data — persistence and time. Managed crawling, entity resolution, historical depth, team alerts. If your problem is "I need to know by Monday," a spreadsheet and the free layers above win. If it's "I need to know within 90 seconds, every day, across 500 sources," buy something or build — and budget for the build like it's a salary, because it is.

I run the free-then-measure path across ~200 public Telegram channels for market and OSINT signals. The collection-layer playbook — cadence design, ?before= cursor paging, repost-ring flags, wording-diff alerting — is the $5 bundle: https://heyuhe.gumroad.com/l/poddr. Real sample brief, pay what you want, $0 fine: https://heyuhe.gumroad.com/l/ruldgm.