Documentation

KEIBIDROP connects two computers and shows each other's shared files as a normal folder. Files stream when read; edits sync back. End-to-end encrypted, no cloud, no account.

Start from what you do

Each line is one situation and the pages that cover it, in the order to read them.

Two computers, first time: how to use it, with a clip per step, share a folder, connect two computers, firewalls and guest WiFi
A NAS or home server as the always-on end: run KeibiDrop on a NAS, edit from it in DaVinci Resolve, share read only
Photos, video, audio: post-production, video and audio projects, image editors, catalogs and databases
Investigations: DFIR, read only with the original timestamps, what is preserved
Datasets and GPU boxes: machine learning, automate with kd
AI agents: the MCP server, the virtual folder for agents, two agents, one folder
Code: git, VS Code, editors
Documents and notes: LibreOffice and Word, Obsidian, conflict files
Scripts and big transfers: script a transfer, kd reference, ffmpeg on remote files

Tutorials

Lessons. Follow the steps in order.

How-to guides

Steps to reach one goal.

Explanation

Background and design decisions.

Reference

Facts and tables.

Common tasks

Terms used in these docs

Term	Meaning
peer	The other computer in a session. A session has exactly two.
fingerprint	The 86-character code that is an installation's identity. The app calls it the peer code.
contact	A saved peer. Reconnects without a new code exchange.
mount, shared folder	The folder where the peer's shared files appear. A virtual filesystem.
conflict file	A sibling file, `name.conflict-<timestamp>.ext`, that preserves the losing version of a simultaneous edit.
relay	The server that passes encrypted pairing data. It cannot read it.
bridge	The server both peers dial out to when a direct connection is impossible. It forwards encrypted bytes only.
room	One two-peer session. One side creates it, the other joins.

For AI agents

Every page here is plain HTML without scripts; curl returns the full content. A machine-readable index of this documentation is at /llms.txt.

To operate KeibiDrop from a script or an agent, use kd. It prints one JSON line per command and exits 0 only on success. Read the kd reference first.

What shipped and what is next: Releases · Roadmap.

Scenario walkthroughs (git repos, databases, media) live on the use cases page. Engineering write-ups are on the blog.