# captchaforge vs. the alternatives
Pick the row that matches your use case.
| Open source | ✅ MIT/Apache | ❌ closed | ❌ SaaS | ❌ SaaS | ❌ SaaS | ✅ MIT |
| Self-hosted | ✅ | partial (extension) | ❌ | ❌ | ❌ | ✅ |
| Cloudflare Turnstile | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ (stealth only) |
| reCAPTCHA v2 | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| reCAPTCHA v3 | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| hCaptcha | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| Akamai BotManager | ✅ | ❌ | ✅ | ✅ | partial | ❌ |
| DataDome | ✅ | ❌ | ✅ | ✅ | partial | ❌ |
| PerimeterX | ✅ | ❌ | ✅ | partial | partial | ❌ |
| GeeTest v3/v4 | ✅ | partial | ✅ | ✅ | ✅ | ❌ |
| AWS WAF | ✅ | ❌ | ✅ | partial | ❌ | ❌ |
| Arkose / FunCaptcha | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| Per-domain learning | ✅ PatternStore | ❌ | ❌ | ❌ | ❌ | ❌ |
| Per-fingerprint learning | ✅ FingerprintLru | ❌ | ❌ | ❌ | ❌ | ❌ |
| Stealth profiles | ✅ 12 named | partial | ❌ | ❌ | ❌ | ✅ |
| JA3/JA4 catalogue | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Network-layer replay | ✅ CF | ❌ | ❌ | ❌ | ❌ | ❌ |
| WASM in-page | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Per-IP rate limit | ✅ | n/a | n/a | n/a | n/a | n/a |
| Prometheus metrics | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Cost per solve | $0 + your infra | extension cost | $1-3 / 1k solves | $0.50-2 / 1k | $1-2 / 1k | $0 (stealth only) |
| Latency | sub-second when chain learns the win | varies | 10-60s (human) | 3-20s | 5-20s | n/a |
| Audit / inspect logic | ✅ open code | ❌ | ❌ | ❌ | ❌ | ✅ |
## When to use which
- **Use captchaforge** when you want self-hosted control, custom
solver logic, predictable cost (your infra only), and an open
audit trail.
- **Use 2captcha / CapSolver / CapMonster** when you need
human-in-the-loop fallback for unsolvable visual puzzles AND
you're OK with per-solve cost + sending traffic to a SaaS.
- **Use puppeteer-extra-stealth** when stealth is all you need
(no CAPTCHA actually fires).
- **Combine captchaforge + a SaaS** for best results: captchaforge
chains its own solvers first, falls through to the SaaS via
`ThirdPartyCaptchaSolver` when its own attempts fail. Lower
total cost + higher coverage than either alone.
## Honest gaps in captchaforge today
- **Network-layer hCaptcha + reCAPTCHA** are stubs — those vendors
fall through to the browser-based chain.
- **Real-world bypass scorecard** is being populated this week
(see https://captchaforge.santh.dev/scorecard).
- **Marketing** is at zero — no users, no GitHub stars, no
word-of-mouth yet. The codebase is real; the community is brand
new.
Numbers shipped here will be refreshed nightly once the scorecard
is live.