Using Settings

Using Settings

Everything you configure through the UI lives in three separate dialogs, each with a clear purpose.

Preferences

Personal settings, for your user.

General — theme (system/light/dark/preset/custom), language, custom system prompt, model fallback order, and custom colors (background, foreground, card, border, primary, accent, muted, sidebar, user bubble color).
Memory — list of persistent memories (key-value pairs). Add, edit inline, delete one or clear all, with a last-updated timeline.
Account — name (editable), email (read-only), role (root/admin/member/viewer).

Integration and extensibility settings, in two tabs.

Integrations — external connectors (manual API key or OAuth — GitHub, GitLab, Google, Slack, etc.) side by side with your custom per-user environment variables (masked in the UI, never exposed in plain text). Shows connected/disconnected status and the webhook URL when applicable.
Provider Routing — local and dynamic LLM providers, e.g. discovering and selecting models served by a local Ollama instance.

Installing MCP connectors and skills from a curated catalog is handled by the workbench’s Library tab, not here — see Using the Workbench.

Instance-wide settings — only visible to administrators.

Users — list with role, created at, last login. Change role, delete a user. Invites section: create an invite (role + optional email + 1–720h TTL), copy URL, revoke pending invites.
Tools — list of global tools by category, with an enable/disable toggle per tool.
Safe Folders — a whitelist of paths that require extra approval even after being trusted.
System — backend version, Python version, platform, status of each service, recent observability span count.
Configuration — allow_public_signup, the instance’s default model, agent max recursion depth, database DSN (read-only), integration token.