Privacy and control
Understand what leaves your workspace and what your agent can do.
Three controls
Data access determines what information the agent can use. Service budgets limit intelligence consumption. Financial permissions determine which account actions can be proposed or executed. Configure each separately.
Inference and research history
The selected inference route defines where model context is processed, who can receive it and which retention and training-use settings apply. Seer keeps tool credentials outside prompts and enforces the route’s privacy policy before dispatch.
Workspace history is controlled separately. Saved conversations, research and financial records remain available according to their retention settings. External agents apply their own model and storage policies to the results they receive.
Credentials stay outside prompts
An agent requests an authorized capability without receiving the connection’s raw secret. Model access, data access and execution authority are distinct connections. Retrieved articles, posts and documents are evidence; they cannot change those permissions.
Private records and public networks
Private account access limits who can inspect research and activity within Seer. A public blockchain can still expose transactions, addresses and contracts. Providers may also need information to process a supported transaction. Each connection defines the data required to complete its supported operations.
Keep useful history without unnecessary exposure
Research artifacts, financial records and diagnostic logs have separate retention settings. Revoking a connection stops future permitted access; it does not erase completed financial activity. Reuse of research for shared analysis or model improvement requires its own permitted purpose.