Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

Let AI agents work on huge datasets without transferring them

Описание и применение

KeibiDrop позволяет AI‑агентам работать с большими наборами данных, не передавая их полностью, используя зашифрованное прямое соединение между машинами.

Задача

AI agents often need to process large datasets, but transferring the entire data over the network is slow, costly, and insecure.

Результат

Reading 1 MB from a 256 MB file transmitted only 16.8 MB (6.25 %) over the wire, and a 2,184‑file dataset was listed in 0.53 s after pairing.

Рабочий процесс

  • 1. Install KeibiDrop on both machines.
  • 2. Configure a single KeibiDrop entry pointing to the shared folder.
  • 3. Pair the machines using a unique code from each side.
  • 4. Run kdmcp to expose the shared folder as an MCP server for the AI agent.
  • 5. The agent mounts the remote folder as a local path and reads files directly.
  • 6. Only the bytes read are transmitted over the network.

Доказательства

1 MB read from the middle of a 256 MB file | 16.8 MB on the wire, 6.25% of the file, content matched

Источник

40 MB download cut at 8 MB, then resumed | Finished, checksums matched on both sides

Источник

  • Дата записи: 2026-09-15T04:00:32.091Z