Dropstone

Settings Guide

Complete guide to configuring Dropstone Agent settings for optimal performance and workflow

This comprehensive guide walks you through all Dropstone Agent settings, helping you configure the agent to match your workflow, security requirements, and preferences.

Accessing Settings

Open Dropstone Agent settings:

  1. Click the More Menu (three dots) in the top navigation
  2. Select Settings (gear icon)
  3. Or use Command Palette: Ctrl+Shift+P → "Dropstone: Open Settings"

1. API Configuration

Configure the AI provider that powers Dropstone Agent.

Selecting Your Provider

Dropstone Agent supports four active API providers:

Dropstone (Default)

Best for: Most users, balanced performance and cost

Setup:

  1. Select Dropstone from provider dropdown
  2. Click Authenticate or Sign In
  3. Complete OAuth flow in browser
  4. Return to Dropstone Desktop - connection confirmed

Features:

  • Native integration with Dropstone infrastructure
  • Optimized performance
  • Automatic model updates
  • Simplified billing

AWS Bedrock

Best for: Enterprise users with AWS infrastructure

Setup:

  1. Select AWS Bedrock from provider dropdown
  2. Enter AWS credentials:
    • Access Key ID
    • Secret Access Key
    • Region (e.g., us-east-1, us-west-2)
  3. Click Test Connection
  4. Select available Claude models

Requirements:

  • Active AWS account
  • IAM permissions for Bedrock
  • Claude model access enabled in AWS console

Configuration Options:

  • Cross-region inference
  • Custom endpoint URLs
  • Session token support

Ollama

Best for: Local inference, privacy-focused users, offline work

Setup:

  1. Install Ollama on your machine: ollama.ai
  2. Pull desired models: 
    ollama pull llama2
ollama pull codellama
  3. In Dropstone settings, select Ollama
  4. Enter Ollama URL (default: http://localhost:11434)
  5. Select from available local models

Advantages:

  • Complete privacy (no data leaves your machine)
  • No API costs
  • Works offline
  • Full control over models

Requirements:

  • 8GB+ RAM (16GB+ recommended)
  • GPU optional but recommended
  • Sufficient disk space for models

Claude Code CLI

Best for: Users with Claude desktop application

Setup:

  1. Install Claude desktop application
  2. In Dropstone settings, select Claude Code CLI
  3. Specify Claude CLI path (auto-detected if installed)
  4. Authenticate if prompted

Features:

  • Integration with Claude desktop
  • Shared conversation history
  • Consistent experience across tools

Model Selection

After configuring your provider, select specific models:

Primary Model

  • Main model for most tasks
  • Recommended: Claude Sonnet 3.5 or equivalent

Fallback Model

  • Used when primary model is unavailable
  • Typically a faster, lighter model

Thinking Models (if available)

  • Extended reasoning for complex problems
  • Higher token usage but better results

2. Model Parameters

Fine-tune AI behavior for your needs.

Temperature (0.0 - 1.0)

Controls output randomness and creativity.

0.0 - 0.3: Deterministic

  • Focused, consistent responses
  • Best for: Bug fixes, refactoring, code review
  • Example: "Fix this function to handle null values"

0.4 - 0.7: Balanced (Recommended)

  • Good mix of consistency and creativity
  • Best for: Feature implementation, general coding
  • Default setting for most users

0.8 - 1.0: Creative

  • More varied and innovative responses
  • Best for: Brainstorming, design, exploration
  • Example: "Suggest creative solutions for this UX problem"

Recommendation: Start at 0.5, adjust based on task type.

Thinking Budget

For thinking-enabled models (extended reasoning).

Options:

  • Default: Standard thinking allocation
  • Conservative: Minimal extended thinking (faster, lower cost)
  • Moderate: Balanced thinking budget
  • Extensive: Maximum thinking for complex problems

When to use Extensive:

  • Complex architecture decisions
  • Multi-step algorithmic problems
  • Novel solution design
  • Deep code analysis

When to use Conservative:

  • Simple tasks
  • Quick fixes
  • Routine operations

Reasoning Effort

Controls how deeply the agent analyzes problems.

Low

  • Fast responses
  • Surface-level analysis
  • Best for: Simple questions, quick fixes

Medium (Recommended)

  • Balanced speed and depth
  • Suitable for most tasks

High

  • Deep analysis
  • Thorough consideration
  • Best for: Complex problems, critical code

3. Auto-Approve Settings

Configure which operations the agent can perform without asking for approval each time.

Understanding Auto-Approve

Security Consideration: Start restrictive, gradually enable auto-approve as you build trust with the agent.

Read Operations

Auto-Approve File Reading

  • ☑️ Enable: Agent reads files automatically when needed
  • ☐ Disable: Agent asks before reading each file

Recommendation: ✅ Enable (safe operation, speeds up workflow)

Auto-Approve Directory Listing

  • ☑️ Enable: Agent explores project structure freely
  • ☐ Disable: Agent asks before listing directory contents

Recommendation: ✅ Enable (read-only operation)

Auto-Approve Code Searches

  • ☑️ Enable: Agent searches codebase without asking
  • ☐ Disable: Agent requests permission for each search

Recommendation: ✅ Enable (read-only, very helpful)

Write Operations

Auto-Approve File Creation

  • ☑️ Enable: Agent creates new files automatically
  • ☐ Disable: Agent asks before creating files

Recommendation: ⚠️ Start Disabled, enable after trust built

Auto-Approve File Modifications

  • ☑️ Enable: Agent edits existing files automatically
  • ☐ Disable: Agent asks before each modification

Recommendation: ⚠️ Start Disabled for critical projects

Auto-Approve File Deletion

  • ☑️ Enable: Agent deletes files automatically
  • ☐ Disable: Agent always asks before deletion

Recommendation: ❌ Keep Disabled (high-risk operation)

Execute Operations

Auto-Approve Command Execution

  • ☑️ Enable: Agent runs terminal commands automatically
  • ☐ Disable: Agent asks before running commands

Recommendation: ❌ Keep Disabled unless working on trusted projects

Allowed Commands Whitelist

If you enable command auto-approval, specify safe commands:

npm run compile:*
npm run build:*
npm run test
git status
git diff

Example Configuration:

{
  "allowedCommands": [
    "npm run compile:*",
    "npm run build:*",
    "npm test",
    "git status",
    "git diff"
  ]
}

4. Browser Settings

Configure browser automation for research and testing.

Viewport Configuration

Width: 1280px (default)

  • Adjust for responsive testing
  • Range: 320px - 3840px

Height: 800px (default)

  • Adjust for long pages
  • Range: 240px - 2160px

Device Emulation

  • Desktop (default)
  • Mobile (iPhone, Android)
  • Tablet (iPad, Android tablet)

Screenshot Quality

Quality: 80 (default)

  • Range: 0-100
  • Higher = better quality, larger files
  • 80 recommended for balance

Format Options:

  • PNG (lossless, larger files)
  • JPEG (compressed, smaller files)
  • WebP (modern, best compression)

Remote Browser

Use Cases:

  • Testing on different machines
  • Shared browser sessions
  • Cloud browser instances

Configuration:

{
  "remoteBrowser": {
    "enabled": true,
    "wsEndpoint": "ws://localhost:9222"
  }
}

5. Checkpoint Settings

Automatic version control for agent changes.

Enable Checkpoints

☑️ Enabled (Recommended)

  • Automatic snapshots before major changes
  • Easy rollback if needed
  • Track agent modifications

☐ Disabled

  • No automatic snapshots
  • Rely on Git for version control only

Checkpoint Frequency

Before every write operation

  • Maximum safety
  • More storage used

Before multi-file operations

  • Balanced approach (recommended)

Manual only

  • You control when checkpoints are created

Maximum Checkpoint History

Default: 10 checkpoints

  • Range: 5 - 50
  • Older checkpoints auto-deleted
  • Consider storage capacity

Storage Location

Default: .dropstone/checkpoints/

Custom Path:

/path/to/custom/checkpoints

Storage Management:

  • Automatic cleanup of old checkpoints
  • Configure max storage size
  • Manual cleanup available

6. Notification Settings

Control audio and voice feedback.

Sound Notifications

Enable Sound Effects: ☑️ / ☐

Volume: 0-100 (default: 70)

Sound Events:

  • ☑️ Task completed
  • ☑️ Approval needed
  • ☑️ Error occurred
  • ☐ Agent thinking (can be distracting)

Sound Selection:

  • Default chime
  • Subtle beep
  • Custom sound file

Text-to-Speech (TTS)

Enable Voice Output: ☑️ / ☐

Use Cases:

  • Accessibility
  • Hands-free coding
  • Long explanations while multitasking

Voice Selection:

  • System default
  • Male/Female options
  • Multiple language support

Speed Control: 0.5x - 2.0x (default: 1.0x)

Volume: 0-100 (default: 70)

What's Spoken:

  • ☑️ Task completion summaries
  • ☑️ Error messages
  • ☐ All agent responses (verbose)

7. Context Management

Control how much information the agent tracks and uses.

Workspace Files

Maximum Open Tabs to Track: 20 (default)

  • Agent monitors open files for context
  • Higher = better context, more memory

File Limit for Context Inclusion: 50 (default)

  • Maximum files added to conversation context
  • Prevents token overflow

Fuzzy Match Threshold: 0.7 (default)

  • Range: 0.0 - 1.0
  • Controls file search sensitivity
  • Lower = more results, less precise

Memory Settings

Context Window Size

  • Small: 8,000 tokens (faster, less context)
  • Medium: 32,000 tokens (balanced)
  • Large: 128,000 tokens (maximum context)
  • Depends on selected model

Memory Retention Duration

  • Session only (clears on close)
  • 1 day
  • 1 week (default)
  • 1 month
  • Forever (until manually cleared)

Automatic Memory Consolidation: ☑️ / ☐

  • Periodically summarizes old conversations
  • Reduces token usage
  • Preserves important patterns

8. Terminal Settings

Configure terminal integration and command execution.

Output Configuration

Maximum Output Lines: 1000 (default)

  • Prevents overwhelming context with logs
  • Range: 100 - 10,000

Command Delay: 300ms (default)

  • Delay between sequential commands
  • Prevents race conditions
  • Range: 0 - 5000ms

Shell Integration Timeout: 5000ms (default)

  • Maximum wait for shell response
  • Range: 1000 - 30000ms

PowerShell Settings (Windows Only)

PowerShell Counter

  • Tracks PowerShell session state
  • Auto-increments for new sessions

Execution Policy Handling

  • Automatically handle restricted policies
  • Temporarily bypass if needed
  • Restore after execution

Recommended Configuration:

{
  "terminal": {
    "maxOutputLines": 1000,
    "commandDelay": 300,
    "timeout": 5000
  }
}

9. Experimental Features

Access cutting-edge capabilities (may be unstable).

Available Experimental Features

☐ Enhanced Pattern Recognition

  • Advanced learning algorithms
  • Beta testing phase

☐ Predictive Code Completion

  • AI-powered suggestions while typing
  • May impact performance

☐ Cross-Project Learning

  • Share patterns across projects
  • Privacy implications

☐ Real-Time Collaboration

  • Multiple users, one agent session
  • Requires shared workspace

Warning: Experimental features may:

  • Cause instability
  • Change without notice
  • Be removed in future versions
  • Have limited support

Recommendation: Enable one at a time, test thoroughly.

10. Language Settings

Configure interface language.

Supported Languages

  • English (default)
  • Spanish (Español)
  • Chinese Simplified (简体中文)
  • Chinese Traditional (繁體中文)
  • Japanese (日本語)
  • Korean (한국어)
  • French (Français)
  • German (Deutsch)
  • Portuguese (Português)
  • Russian (Русский)

Note: Language change requires Dropstone Desktop restart.

11. Privacy & Data

Control what data is collected and stored.

Telemetry

☑️ Enable anonymous usage statistics

  • Helps improve Dropstone Agent
  • No personal code or data collected
  • Only feature usage metrics

☐ Disable all telemetry

  • Complete privacy
  • No usage data sent

Local Storage

Where data is stored:

  • Windows: %APPDATA%\Dropstone\
  • macOS: ~/Library/Application Support/Dropstone/

What's stored locally:

  • Agent conversation history
  • Learned patterns
  • Settings and preferences
  • Checkpoints

Data Control:

  • Export conversation history
  • Clear all local data
  • Delete specific conversations

Workspace Collaboration

☐ Enable workspace sync (optional)

  • Share history with team
  • Collaborative learning
  • Requires explicit opt-in

12. Performance Settings

Optimize agent performance for your machine.

Resource Limits

Maximum Memory Usage: 4GB (default)

  • Range: 2GB - 16GB
  • Adjust based on available RAM

Maximum CPU Cores: 4 (default)

  • Range: 1 - all available
  • Higher = faster but more resource intensive

Background Processing: ☑️ / ☐

  • Agent continues working in background
  • May slow down other applications

Optimization Options

☑️ Enable caching

  • Faster repeat operations
  • Uses more disk space

☑️ Lazy loading

  • Load features on demand
  • Faster startup time

☐ Aggressive pre-loading

  • Load everything at startup
  • Slower start, faster operation

Configuration 1: Beginner / Cautious

Best for: New users, critical projects

API Provider: Dropstone
Temperature: 0.5
Auto-Approve Reads: ✅
Auto-Approve Writes: ❌
Auto-Approve Commands: ❌
Checkpoints: ✅ (every operation)
Notifications: ✅

Configuration 2: Intermediate / Balanced

Best for: Most users, daily development

API Provider: Dropstone or AWS Bedrock
Temperature: 0.5
Auto-Approve Reads: ✅
Auto-Approve Writes: ✅
Auto-Approve Commands: ❌
Allowed Commands: build, test, compile
Checkpoints: ✅ (multi-file operations)
Notifications: ✅

Configuration 3: Advanced / Autonomous

Best for: Experienced users, trusted projects

API Provider: Dropstone, AWS, or Ollama
Temperature: 0.6
Auto-Approve Reads: ✅
Auto-Approve Writes: ✅
Auto-Approve Commands: ✅ (whitelist)
Allowed Commands: [comprehensive list]
Checkpoints: ✅ (smart frequency)
Experimental Features: Some enabled

Configuration 4: Privacy-Focused

Best for: Sensitive projects, privacy requirements

API Provider: Ollama (local)
Telemetry: ❌
Workspace Sync: ❌
Auto-Approve Writes: ❌
Checkpoints: ✅
Local storage only

Backing Up Settings

Export Settings

  1. Open Settings
  2. Click Advanced → Export Settings
  3. Save JSON file
  4. Store securely

Import Settings

  1. Open Settings
  2. Click Advanced → Import Settings
  3. Select JSON file
  4. Review changes
  5. Apply

Settings File Location

Direct access:

  • Windows: %APPDATA%\Dropstone\config.json
  • macOS: ~/Library/Application Support/Dropstone/config.json

Manual editing:

  • Open file in text editor
  • Modify JSON carefully
  • Restart Dropstone Desktop

Troubleshooting Settings

Settings Not Saving

Solution:

  1. Check file permissions
  2. Ensure Dropstone has write access
  3. Try running as administrator (Windows)
  4. Check disk space

Settings Reset After Update

Solution:

  1. Export settings before updating
  2. Re-import after update
  3. Report issue if persistent

Can't Connect to Provider

Solution:

  1. Verify API credentials
  2. Check internet connection
  3. Test connection in settings
  4. Review firewall rules
  5. Check provider status page

Next Steps

Configure settings once, adjust as you learn. The agent adapts to your preferences over time.