OpenDoor helps people find open source projects where newcomers actually get in. “Good first issue” lists show what a project says; OpenDoor shows what it does. From the dependency graphs of npm, PyPI and Go (deps.dev) plus well-known GitHub apps, we analyzed 958 projects with monthly GitHub activity since 2018 and the OpenSSF Scorecard. Business rules give each project a profile and a “ready for beginners” badge, which it only gets if the data proves five things: outsiders had their first PR merged in the last year, at least 1 in 4 new contributors got a PR accepted, outside PRs were merged in the last 3 months, outside PRs get a reply within 14 days, and there is a contribution guide. You pick your level, languages and how you want to help (code, docs, triage, review, translation); OpenDoor matches you with projects and shows where to start. An LLM (Gemma 4) reads each project’s contribution guide to classify its AI policy with a literal quote, suggest where help is needed and write a responsible “where to start” guide: observe, talk to maintainers, then contribute. Everything is generated offline and cached; the site is static and fast.