Eyeban

Configuration

Settings sections, official cloud endpoint, permissions, sandbox, privacy modes, and Local API.

Eyeban exposes user and project settings for editor behavior, agent defaults, permissions, privacy, network, and integrations. Prefer checking shared agent guidance into the repository; never commit secrets.

Settings you will use often

Typical Settings areas include:

  • Account and sign-in
  • Agent defaults (mode, model preferences where entitled)
  • Bring your own key (optional provider keys under Agents)
  • Permissions (Ask / Allow safe / Allow all)
  • Shell sandbox (Off / Soft / Restricted)
  • Privacy mode
  • Official server / cloud endpoint
  • Remote / SSH related options
  • Local API
  • Network diagnostics
  • Editor and workbench preferences

Official server

Keep the official server toggle pointed at api.eyeban.ir for production use. Remote workspaces should use HTTPS to the cloud endpoint. Do not point a production workflow at unofficial mirrors.

Permissions and shell sandbox

Permissions gate tool execution. Shell sandbox levels:

  • Off — no extra sandbox layer from this setting
  • Soft — default balanced restrictions for routine agent shells
  • Restricted — tighter limits when you want more containment

Pair sandbox choice with permission level; Restricted plus Ask is a cautious default for unfamiliar repos.

Privacy mode

Privacy mode values include:

  • off
  • no_training
  • no_storage

See Privacy in the app and the Legal pages for binding policy language.

Local API

The workbench can expose a Local API on 127.0.0.1:17300 for machine-local automation. It is not a public internet service. See the API page for endpoint groups.

Diagnostics

Use Network diagnostics in Settings when agent runs fail to connect, models do not load, or SSH remote sessions misbehave.