MSN Messenger AI — Resurrected

🪟 It's 2005 again — but your buddies got smarter

MSN Messenger died in 2014. Everyone remembers the nudges, the emoticons, the dramatic away messages, the buddy list full of xX_names_Xx. MSN Messenger AI — Resurrected brings the whole experience back in the browser, with one twist: your contacts are AI personas powered by Gemini.

What it does

Full MSN Messenger 2005 UI — buddy list with online contacts, display pics, the classic sign-in screen ("It's 2005 again — but the bots got smarter"), title bar, status text, timestamps.
3 AI contacts with real personalities — Gemini_Assistant (helpful, light MSN-era flavor), Comedian_Bot_2005 (lol/brb/roflcopter humor), xX_Dramaqueen_Xx (extremely dramatic 2005 emo energy). Each has its own system prompt and its own multi-turn chat session — conversations remember context.
Nudge — the window physically shakes, and the bot reacts. Obviously.
Group chat — all AI contacts in one conversation, and the drama writes itself.
Status selector — Online / Busy / Away / Be Right Back with colored indicators; your AI buddies react to your status like real 2005 buddies would ("brb? oki cya l8r").
Save conversation — download any chat as a .txt file, exactly like the old "Save conversation" feature.
Display pictures — cycle through emoji display pics like it's 2005.

How it's built

Vanilla JS + esbuild, styled as pixel-era MSN. Chat runs on Firebase AI Logic against the Gemini Developer API backend — cloud Gemini with per-bot system instructions and true multi-turn sessions (startChat per persona). Firebase App Check protects the AI backend. Hosted fully static on GitHub Pages.

🔧 Built with Kiro

This project is built with Kiro. The /.kiro directory at the repo root is the proof:

.kiro/steering/ — product, tech and structure steering docs that keep every change aligned with the Resurrection vision.
.kiro/specs/msn-ai-resurrection.md — the original resurrection spec.
.kiro/specs/msn-status-export/ — full spec-driven development (requirements → design → tasks) for the status-selector and save-conversation features, implemented by the Kiro CLI agent end-to-end: spec written, features implemented, esbuild bundle rebuilt and verified — all inside a single Kiro CLI session.

Kiro did the heavy lifting on the second feature wave: the steering docs, the spec (requirements/design/tasks), the implementation of both features, the bundle rebuild and the verification grep — one agentic run, spec-driven, no manual coding.