Private by default,
with controls you can see.
Where your keys live, what leaves your Mac, and where code runs — stated plainly, with nothing we can't point to.
Six things that hold
How your work is handled.
Keys stay in your Keychain
Plugin tokens live in the macOS Keychain as connector.<name> — never in the database, never in a file. The database keeps ids and tool definitions, never secrets.
One guarded client
Plugin traffic goes from your Mac to the provider through one guarded client. Its rules live in the engine, not in a settings page.
Tools you approved, nothing else
A tool reaches the model only after you approve it at its current definition, and approved tools are offered only in the runs you switch plugins on for.
Only what the task needs
The context engine sends whole definitions or a labelled signature, and lists every segment in a manifest with its reason and size.
Runs happen in a sandbox
Agent runs use a sandbox — locally, or in a cloud sandbox that only your account can see. Cloud sandboxes are never public and stop on their own after 15 minutes.
Games can't reach the network
Play runs in a cookie-free sandbox with no network access.
A manifest, not a mystery.
Before a Build run, the engine picks the definitions the task needs and writes down why. You can read the list the model was given.
How the context engine works →Security testing
For surfaces you're authorized to test.
The Service includes authorized security testing features — recon, port and web scans with one report. Use them only on systems you own or are authorized to assess, under the Terms of Service.
Results are part of the content covered by the Privacy Policy, which lists “authorized security testing results” among the data handled in connection with the Service.