Give your AI agents a shared private brain across all your machines.
Five agents. Five machines. One brain.
Instead of every agent starting from scratch, your fleet compounds context through one private machine network.
Agents remember what matters — as plain markdown.
Memories live as human-readable markdown in your Obsidian vault, sync across trusted machines through Hivemind Sync, and are retrieved through a local API — never a hosted memory service.
Every memory knows where it came from.
but which agent, on which machine, inside which part of your private Hivemind.
Local recall. Fleet-wide memory.
Each trusted machine keeps its own local copy of the brain. Agents get fast local recall while shared memory propagates across the fleet.
One local endpoint. Four verbs.
No hosted vector database. No central SaaS brain. No cloud memory vendor required.
Memory that can prove itself.
Prove what the fleet remembered, when, and which agent wrote it — without exposing the raw memory body.
The memory layer never leaves your vault.
Markdown notes, retrieval index, and proof receipts all live in your private shared brain — no hosted vector database.
Rich recall in milliseconds.
| Memories | Recall p50 | Recall p95 | Answer p50 |
|---|---|---|---|
| 100 | 0ms | 0ms | 0ms |
| 500 | 0ms | 0ms | 0ms |
| 1500 | 0ms | 0ms | 0ms |
First indexed reads were effectively as fast as cached reads. The private local index keeps rich, typed recall in milliseconds — versus hundreds of milliseconds scanning raw markdown.
HivemindOS · Private by default