# Custom Modes Guide
Learn how to create custom observation modes for different workflows.
## What are Modes?
Modes define how claude-recall captures and categorizes observations. The default `code` mode is optimized for software development, but you can create custom modes for:
- Research and analysis
- Writing and documentation
- Data exploration
- System administration
## Mode Structure
A mode is a JSON file in `extension/profiles/`:
```json
{
"id": "research",
"name": "Research Mode",
"description": "Optimized for research and analysis workflows",
"observation_types": [
{
"type": "finding",
"description": "A research finding or insight",
"icon": "🔍"
},
{
"type": "source",
"description": "A source or reference",
"icon": "📚"
},
{
"type": "hypothesis",
"description": "A hypothesis to test",
"icon": "💡"
}
],
"concepts": [
"methodology",
"data-source",
"conclusion",
"limitation"
],
"prompts": {
"observation": "Analyze this interaction and extract research-relevant observations...",
"summary": "Summarize this research session..."
}
}
```
## Creating a Custom Mode
### 1. Create the mode file
```bash
cp extension/profiles/code.json extension/profiles/my-mode.json
```
### 2. Edit the configuration
```json
{
"id": "my-mode",
"name": "My Custom Mode",
"observation_types": [
// Define your observation types
]
}
```
### 3. Activate the mode
In `~/.claude-recall/settings.json`:
```json
{
"CLAUDE_RECALL_MODE": "my-mode"
}
```
### 4. Restart the worker
```bash
claude-recall restart
```
## Example: Writing Mode
```json
{
"id": "writing",
"name": "Writing Mode",
"description": "For creative writing and documentation",
"observation_types": [
{
"type": "draft",
"description": "A draft or revision",
"icon": "📝"
},
{
"type": "feedback",
"description": "Feedback or critique",
"icon": "💬"
},
{
"type": "reference",
"description": "A reference or inspiration",
"icon": "🔗"
}
],
"concepts": [
"tone",
"structure",
"audience",
"style"
]
}
```
## Community Modes
Check out community-contributed modes in `contrib/modes/`:
- `research.json` - Research and analysis
- `devops.json` - System administration
- `data.json` - Data science workflows
## Tips
1. **Keep observation types focused** - 3-5 types is usually enough
2. **Use descriptive concepts** - They help with search and filtering
3. **Customize prompts** - Tailor the AI prompts to your workflow
4. **Test incrementally** - Start with small changes and iterate
