# Clone Flow

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Melissa Rajamanuvel](https://cerebralvalley.ai/u/Melissarine)
- **GitHub:** https://github.com/Datellx/CloneFlow
- **Demo video:** https://vimeo.com/1243028020
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1194

The problem: When someone leaves a company, everything they know leaves with them. The next person spends days or weeks figuring out what the last person already knew.

What Clone Flow does: It captures what a person knows before they leave, and hands it to the next person right where they work.

How it works:

The company starts a handover for the person who is leaving, in plain words.
The person connects their work accounts. Gmail, Outlook (mail and meetings) and Teams, one click each, secure and read-only.
Clone Flow separates personal from professional. Personal messages are set aside in code, before any AI sees them.
It maps everything they know into a knowledge base: every piece of knowledge saved with proof of where it came from.
It finds the gaps and fills them. Clone Flow interviews the person about what was never written down. They answer by talking or typing, or share their screen to show where things are.
It goes beyond the desk, onto the production floor. Much of the most valuable know-how lives in people’s hands: calibrating a machine, fixing a part, running a line. The person can film the work or walk the floor with their phone. Meta’s SAM 3.1 computer vision tracks the equipment and their hands frame by frame, and Clone Flow turns it into step-by-step instructions, each linked to the exact moment in the video.
It delivers the knowledge to the right person. Clone Flow connects to Microsoft Teams, so the next person just asks: “@Clone Flow, how do I fix the upside-down wrist on a new arm?” They get the answer in the original person’s own words, with proof. If nobody ever covered it, Clone Flow says “I don’t know” instead of guessing.

Built with Meta’s models: Muse Spark (reads, maps knowledge, finds gaps, interviews and answers), Muse Voice Transcribe (hears the person) and SAM 3.1 (computer vision for hands-on work).

Proven on a real handover: 626 messages read, 9 of 9 personal emails kept private, answers in Teams in about 4 seconds, and $0.93 of Muse model calls for the whole handover.

Built from scratch during the hackathon (Oct 26 – Nov 4, 2026).

## More from Meta Global AI Developer Hackathon

- [GODMODE](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1191)
- [Adrian Gajewski](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1192)
- [Gatekeeper](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1193)
- [Furkan Kupcu](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1195)
- [Max Wo](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1196)
- [Listing Partner](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1197)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1194. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt