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

Local or hosted agent memory: what actually leaves your machine

Description and use

The company clarified that memory storage stays on the chosen device, while processing occurs transiently on a secure hub in Zurich with no data retention. Users can verify this by inspecting the source‑available client and network traffic.

Problem

Users cannot verify whether their data truly stays on their machine or leaves to a vendor’s servers, because privacy claims are vague and untestable.

Outcome

The company clarified that storage never leaves the chosen device, while processing is transient and not retained. Users can verify this by inspecting the client code and network traffic.

Workflow

  • 1. User signs up and selects local or hosted edition.
  • 2. Entire memory (curated records, chat archive, graph, notes, playbooks, search index) is stored in a container on the user’s machine (local) or on the vendor’s VPS (hosted).
  • 3. When a conversation needs curation, a window of exchanges is sent to the hub for processing.
  • 4. The hub processes the data using a zero‑data‑retention model and returns results; no data is stored by the hub beyond rules and project lines.
  • 5. The hub runs in a Tier III+ data center in Zurich, Switzerland, with a zero‑data‑retention contract.
  • 6. The user can verify the data flow by inspecting the source‑available client code and observing network traffic.
  • 7. If the hub is unreachable, local memory continues recording and searching; only brief, recall, and tidying wait until the hub returns.

Evidence

"nothing of your memory ever leaves your machine."

Source

The hub is a secure compute instance in Zurich, Switzerland, in a Tier III+ datacenter — redundant power, generator backup, the tier as the vendor states it in its own security documentation.

Source

The entire memory: the curated records, the full chat archive, the graph of decisions, the notes, the playbooks, and the search over all of it. It lives in a container on your disk — Docker or Podman, roughly 3 GB including the memory model.

Source

The client is source‑available — you can read every line and see exactly what leaves your machine: exchanges go up, the processing asks and search answers go up, commands and result rows come down.

Source

  • Record date: 2026-10-22T19:11:34.122Z