Skip to main content
The preferences file stores personal settings that customize how the agent interacts with you.

​
Location

~/.craft-agent/preferences.json

​
Structure

{
  "name": "Alex",
  "timezone": "America/New_York",
  "location": {
    "city": "New York",
    "region": "New York",
    "country": "USA"
  },
  "language": "English",
  "notes": "I prefer concise responses",
  "updatedAt": 1737451800000
}

​
Fields

​
name

Your name or how you’d like to be addressed. 
"name": "Alex"
The agent may use this in greetings or personalized responses.

​
timezone

Your timezone in IANA format. 
"timezone": "America/New_York"
Used for:
  • Accurate date/time references (“today”, “this week”)
  • Daily note access
  • Task due dates
Common values:
  • America/New_York
  • America/Los_Angeles
  • Europe/London
  • Europe/Berlin
  • Asia/Tokyo
  • Australia/Sydney

​
location

Your location, stored as a nested object with city, region, and country fields: 
"location": {
  "city": "New York",
  "region": "New York",
  "country": "USA"
}
The location fields are used for location-aware responses like weather, local events, or time-sensitive information.

​
updatedAt

Timestamp (in milliseconds) of when preferences were last updated. This is set automatically. 
"updatedAt": 1737451800000

​
language

Your preferred language for responses. 
"language": "English"
The agent will respond in this language when possible.

​
notes

Free-form notes about your preferences. 
"notes": "I prefer bullet points over paragraphs. Technical background."
These notes are included in the agent’s context and influence response style.

​
Editing Preferences

​
Using the Command

> /prefs
Opens an interactive editor for your preferences.

​
Manual Editing

Edit the file directly: 
nano ~/.craft-agent/preferences.json
Changes take effect on the next conversation (start a new session or restart the app).

​
How Preferences Are Used

Preferences are included in the system prompt, so the agent knows:
  • What to call you
  • Your current time and timezone
  • Where you’re located
  • How to communicate with you
This context makes responses more relevant and personalized.

​
Example

{
  "name": "Jordan",
  "timezone": "Europe/Berlin",
  "location": {
    "city": "Berlin",
    "region": "Berlin",
    "country": "Germany"
  },
  "language": "English",
  "notes": "Software engineer. Prefer code examples over descriptions. Use metric units."
}
With these preferences, the agent will:
  • Use Berlin time for scheduling references
  • Respond in English
  • Lean toward showing code
  • Use kilometers, Celsius, etc.