Privacy and data flow

Stellaris Companion is local-first, but some features can use an AI provider you select. The boundary depends on which workflow you use.

The boundary at a glance

flowchart LR
  accTitle: Stellaris Companion privacy and data flow
  accDescr: Raw saves and the parsed archive remain on the player's computer. Only extracted context crosses the boundary when the player uses a selected AI provider or a connected MCP client.
  subgraph Device["Your computer"]
    Save["Raw .sav file"] --> Archive["Parsed campaign archive"]
    Archive --> Advisor["In-app Advisor or Chronicle"]
    Archive --> MCP["Local MCP server"]
  end
  Advisor -->|"Question + extracted context"| Provider["Your selected AI provider"]
  MCP -->|"Structured briefing"| Client["Connected AI client"]
  Client -->|"Depends on client settings"| ClientModel["Client's model provider"]

Always local

• Raw .sav files and the parsed campaign archive.
• Save discovery, extraction, history comparison, and cached Chronicle text.
• The local MCP server and its connection to a desktop AI client.
• Provider credentials, which are not included in MCP responses.

Sent to your selected AI provider

When you use the in-app Advisor or generate Chronicle prose, the selected provider receives the question or writing instruction plus extracted campaign context. It does not need the raw save file.

Local models can keep that inference on your machine. Remote providers receive the submitted context according to their own terms and privacy policies.

Shared with a connected MCP client

An MCP-capable app can request structured briefings, recent events, and cached Chronicle material from the local archive. The client performs the reasoning in its own process and may itself use a remote model. Review that client’s privacy settings separately.

Chronicle write-back tools change only the local Chronicle cache, and agents should use them only after explicit save, apply, or send-back intent. They cannot mutate a Stellaris save.

For the public website’s separate data practices, read the website privacy notice.