Pomerado-style "any website becomes an agent…

Verified: GitHub 1,033 stars (API) / 1.0k (page), +199 on 25 Oct and +683 on 26 Oct (OSS Insight); YC Fall 2026 company page, team of 2; public pricing Free $0 (100 tool calls/day), Premium $50/month (1,000 calls/day), Business custom (pomerado.ai). Not found: revenue, paying-customer count, user count. Claims "2–4x faster, 10–1000x cheaper, +30% more reliable than browser use" are Self-reported. Passes: benchmark Pomerado (YC Fall 2026) is 22 days old with 1,033 GitHub stars, accelerating (+683 on 26 Oct, the peak and last full day; Verified), public paid plan $50/month (Verified, pomerado.ai pricing page), X 439 posts/7d rising from 3–4/day to 284 on 26 Oct (Verified). Weaknesses: the launch spike is only 2 days old (no proof it holds); no public revenue or user count (only stars and a price list); the space is crowded (Browser Use, Browserbase, Apify, Firecrawl, Composio, and Pomerado itself is free and open source under MIT); acting on logged-in LatAm portals (gov.br, banks, tax and invoice portals) needs custody of user credentials and 2FA, which clashes with the HBV rule of no security promises and with LGPD and site terms of use; POS 57. Stars: Total: 1.0k (github.com page text, public HTML read with curl 27 Oct 2026 ~07:03 BRT; the box browser was not used); gh API cross-check 1,033 at 07:02 BRT. Created 2026-10-05 (API only), MIT (page). 24h (26 Oct, last full UTC day): +683. 7d (20–26 Oct): +886 ≈127/day; prior 7d (13–19 Oct): +28; 7v7 = +3,064% (accelerating). 3v3 (24–26 Oct 884 vs 21–23 Oct 2) = accelerating. Since creation: 1,033 in 22 days ≈47/day, but 85% of all stars arrived on 25–26 Oct (YC launch). Peak day 26 Oct (+683, last full day) → not past peak. Daily series (OSS Insight analyze-recent-stars, fetched 27 Oct 06:57 BRT): 20 Oct 0 · 21 Oct 0 · 22 Oct 1 · 23 Oct 1 · 24 Oct 2 · 25 Oct 199 · 26 Oct 683 · 27 Oct partial 118. Not on github.com/trending daily or weekly at 06:58 BRT. All star numbers Verified (page text + API/OSS Insight); nothing estimated. X: 439 posts/7d (Rising +6100% 3v3 (24–26 Oct 372 vs 21–23 Oct 6; daily full UTC days 4, 2, 0, 3, 85, 284); peak 26 Oct 284; 26 Oct is 0% below peak; window 2026-10-20 10:30 UTC → 2026-10-27 09:50 UTC (X API recent counts, granularity day, pulled 27 Oct 2026 ~07:00 BRT; Mon 20 and Tue 27 are partial UTC days; full days Wed 21 → Mon 26; 3v3 = (24+25+26) vs (21+22+23); 24–25 Oct are a weekend)). Revenue hypothesis: ~$1K–$6K/month at month 6 (HYPOTHESIS).

call tools) for any website from a single prompt, hosts it, stores logins with automatic 2FA and repairs the integration when the site changes (YC Launch page and pomerado.ai, read 27 Oct 2026). Why replicable: The core harness is MIT-licensed, so the build is a localisation and hosting job; the hard parts are keeping portal integrations alive and handling user logins safely. Monetization: Usage-based subscription: free tier, then per-tool-call plans in BRL/MXN; paid "portal packs" (tax, invoices, delivery) for accountants (Hypothesis; no approved price). Build: ~3–4 agent-weeks on top of the MIT harness, plus ongoing upkeep of each portal integration; stack: Pomerado MIT core (TypeScript) on Vercel/Fly; Postgres; Stripe + Pix; MCP server. GTM: SEO/GEO pages per portal ("MCP do e-CAC", "agente que emite nota fiscal"), listings in MCP directories, accountant communities.

26 company page, team of 2; public pricing Free $0 (100 tool calls/day), Premium $50/month (1,000 calls/day), Business custom (pomerado.ai). Not found: revenue, paying-customer count, user count. Claims "2–4x faster, 10–1000x cheaper, +30% more reliable than browser use" are Self-reported.