Settings

The Settings page (/settings) is organized as a tabbed interface with a 192px sidebar on the left listing six tabs: Agent, Spending, AI Providers, Memory, Server, and Updates.

Changes on the Agent, Spending, Memory, and Server tabs require clicking Save. A Reset button appears when there are unsaved changes.

Agent Tab

Controls how the AI agent behaves during conversations.

Max Turns per Run

A slider (range 1--200, default 25) setting the maximum number of agent turns (tool calls plus responses) in a single run. Higher values allow longer autonomous workflows. An input field allows entering an exact value.

Temperature

A slider (range 0.0--2.0, step 0.05, default 0.7) controlling randomness. Lower values produce more deterministic output; higher values produce more varied responses.

Max Output Tokens

An optional limit on tokens per response, toggled by a switch. When enabled, a slider (range 256--131072, step 256) and quick-select presets (1K, 4K, 8K, 16K, 32K, 64K, 128K) let you set the value.

Spending Tab

Tracks token usage and provides spending controls.

Current Usage

An overview card showing total tokens (with input/output breakdown), estimated cost in USD, and API call count for a selectable period (daily, weekly, or monthly).

Spending Limits

A toggleable section with:

Reset Period -- daily, weekly, or monthly. Usage counters reset at the start of each period.
Token Limit -- optional. When enabled, shows a progress bar with current usage versus the limit. A slider (100K--50M) and input field set the cap.
Cost Limit (USD) -- optional. When enabled, shows a progress bar and slider ($0.50--$500) with quick-select presets ($1, $5, $10, $25, $50, $100). The agent is blocked when limits are reached.

Progress bars change color from green through yellow, orange, and red as usage approaches the limit.

AI Providers Tab

Manages connections to LLM providers.

Provider Grid

Eight providers displayed as clickable cards in a 4-column grid: Anthropic, OpenAI, Google, Moonshot, Grok, OpenRouter, MiniMax, and Ollama. A green checkmark indicates a configured provider. Click a card to select it.

Provider Detail

When a provider is selected, a card appears with:

API Key / Host URL -- enter the key (password field) or URL (for Ollama). If already configured, you can update or disconnect it.
Test connection -- sends a test request and displays the result.
Models section -- shown for configured providers. A dropdown lists available models that can be enabled. Each enabled model appears in a list with a remove button. You can also add custom models by specifying an ID, display name, and per-1M-token pricing.

Setting a provider as default and selecting a default model is done from the chat input's model selector.

Memory Tab

Configures the embedding system used for workspace memory search.

Embedding Provider -- select which AI provider handles embedding generation.
Embedding Model -- select the specific embedding model (options depend on the chosen provider).
Test -- optionally provide an API key and run a test embedding to verify dimensions and connectivity.

Server Tab

Network configuration for the gateway:

Port -- the HTTP/WebSocket port (default 3777).
Host -- the bind address (default 127.0.0.1).

Changes require a server restart to take effect.

Updates Tab

Manage application updates. Behavior differs between desktop and web modes.

Desktop Mode

Check for updates -- queries the update server for a new version.
Download -- if an update is available, downloads it with a progress indicator.
Install and restart -- applies the downloaded update and restarts the app.

Web Mode

Check for updates -- queries the API for the latest available version.
Displays the current version and the latest version with a comparison.

Status indicators show the current state: idle, checking, available, up-to-date, downloading (with percentage), downloaded, or error.