Execute Any MCP Tool Directly

The Tool Execution screen gives you unlimited access to all MCP tools from every connected server. Execute any tool manually through the UI with rich parameter configuration and comprehensive result display - no limits, no restrictions.

Unlimited Tool Access

Direct Execution Interface

Unlike voice conversations (limited to 20 tools), the Tool Execution screen provides:

• Access to ALL tools from ALL connected MCP servers
• No quantity limits - hundreds of tools available
• Manual parameter control with dynamic forms
• Rich result visualization through the MCP protocol
• Full tool capabilities without restrictions

How It Works

When you connect MCP servers:

1. All tools become available in the Tools screen
2. Select any tool to open Tool Execution
3. Configure parameters through dynamic UI
4. Execute through MCP with real-time feedback
5. View rich results formatted for mobile

Key Features

• Dynamic Form Generation: UI adapts to each tool's parameters
• Real-time Validation: Immediate feedback on parameter values
• Execution Monitoring: Live progress and status updates
• Result Visualization: Structured display of tool outputs
• Error Handling: Clear error messages and recovery options

Execution Flow

1. Tool Selection

Tools reach the execution screen through:

• Voice Command: AI extracts tool and parameters from speech
• Manual Selection: Choose from Tools screen or suggestions
• Conversation Context: AI determines tool needs from chat
• Quick Actions: Shortcuts and saved executions

2. Parameter Configuration

The screen dynamically generates forms based on tool requirements:

GitHub: Create Issue
- Repository* [dropdown of your repos]
- Title* [text input]
- Body [multi-line text]
- Labels [tag selector]
- Assignees [user selector]

3. Execution Process

1. Review Parameters: Verify all inputs before execution
2. Initiate Execution: Tap execute or confirm voice command
3. Monitor Progress: Real-time status updates
4. View Results: Structured output display
5. Take Action: Follow-up options based on results

Parameter Types

Common Parameter Types

Text Inputs

• Single line: Names, titles, IDs
• Multi-line: Descriptions, comments, code
• Formatted: URLs, emails, patterns

Selections

• Dropdown: Choose from predefined options
• Multi-select: Pick multiple values
• Radio buttons: Exclusive choices
• Toggles: Boolean on/off values

Complex Types

• Date/Time: Calendar and time pickers
• File paths: Browse and select files
• JSON: Structured data editor
• Key-Value: Dynamic pair lists

Parameter Validation

Real-time validation ensures correct inputs:

• Required Fields: Marked with asterisks
• Format Checking: Email, URL, pattern matching
• Range Validation: Min/max for numbers and dates
• Dependency Rules: Show/hide based on other values
• Custom Validation: Tool-specific requirements

Voice Parameter Extraction

Natural Language Processing

The AI extracts parameters from voice commands:

"Create a GitHub issue titled 'Fix login bug' in the main repo"

Extracted:
- Tool: GitHub Create Issue
- Title: "Fix login bug"  
- Repository: "main" (matched to actual repo name)

Confirmation Flow

For voice-initiated executions:

1. Parameter Display: Shows extracted values
2. Edit Option: Modify before execution
3. Voice Confirmation: "Execute" or "Cancel"
4. Visual Confirmation: Tap to proceed

Execution Interface

Pre-Execution View

• Tool Header: Name and description
• Parameter Summary: All configured values
• Warnings: Any potential issues
• Cost Indicator: If tool has usage costs
• Execute Button: Primary action

During Execution

Real-time updates include:

• Progress Indicator: Spinner or progress bar
• Status Messages: Current operation step
• Partial Results: Data as it arrives
• Cancel Option: Stop long-running operations
• Time Elapsed: Execution duration

Post-Execution View

Results are displayed with:

• Success/Error Status: Clear visual indicator
• Structured Output: Formatted result data
• Raw Response: Optional JSON/text view
• Action Buttons: Follow-up operations
• Share Options: Export or send results

MCP Protocol Integration

Rich Results Display

All tool executions return results through the MCP protocol, displayed with:

• Structured data visualization optimized for mobile
• Interactive elements for follow-up actions
• Formatted output based on response type
• Native mobile controls for navigation
• Export capabilities for sharing results

Tool-Specific Examples

GitHub: Create Pull Request

Parameters:
- Repository: frontend-app
- Title: "Add dark mode support"
- Head Branch: feature/dark-mode
- Base Branch: main
- Body: "Implements #123..."

MCP Execution:
✓ Sending request via MCP protocol
✓ Server processing through GitHub API
✓ Receiving structured response

Rich Result Display:
PR #456 created successfully
- Interactive PR preview card
- Tappable links to GitHub
- Quick actions: Review, Merge, Share
- Related data: Commits, Files changed

Sentry: Query Errors

Parameters:
- Project: production-api
- Time Range: Last 24 hours
- Severity: Error, Critical
- Limit: 50

MCP Execution:
✓ Querying through MCP server
✓ Sentry API processing
✓ Structured error data returned

Rich Result Display:
Found 23 errors with interactive cards:
- Expandable stack traces
- User impact metrics
- Visual error timeline
- Quick assign/resolve actions

Reddit: Search Posts

Parameters:
- Subreddit: programming
- Query: "rust performance"
- Sort: Top
- Time: This week

MCP Execution:
✓ Reddit MCP server processing
✓ API search and filtering
✓ Rich content retrieval

Rich Result Display:
12 posts with native formatting:
- Threaded comment preview
- Upvote/award indicators
- Media previews inline
- Save and share actions

Access Without Limits

Remember: While voice conversations are limited to 20 tools for optimal AI performance, the Tool Execution screen gives you:

• ✅ Unlimited tool access - Use every tool from every server
• ✅ Full MCP capabilities - All features available
• ✅ Rich mobile interface - Optimized for touch
• ✅ Same subscription - No premium tiers needed

Advanced Features

Parameter Templates

Save common parameter combinations:

1. Configure parameters for a tool
2. Save as template with descriptive name
3. Quick load for future executions
4. Share templates with team

Batch Execution

Execute tools multiple times with different parameters:

1. Enable batch mode for supported tools
2. Configure parameter sets (CSV, list, etc.)
3. Review all executions before starting
4. Monitor progress of batch operation
5. Aggregate results in summary view

Conditional Execution

Chain tool executions based on results:

If GitHub PR creation succeeds:
  → Post to team channel
  → Update project board
  → Start CI build

Execution History

Recent executions are saved with:

• Full parameter sets
• Execution results
• Timestamps and duration
• Quick re-run option

Error Handling

Common Error Types

Validation Errors

• Missing required parameters
• Invalid parameter formats
• Out-of-range values

Execution Errors

• Authentication failures
• Network timeouts
• Rate limit exceeded
• Server errors

Permission Errors

• Insufficient access rights
• Resource not found
• Operation not allowed

Error Recovery

Each error provides:

• Clear error message: What went wrong
• Suggested fixes: How to resolve
• Retry options: Quick retry with modifications
• Help links: Detailed documentation

Best Practices

Efficient Execution

1. Use templates: For repeated operations
2. Batch when possible: Reduce round trips
3. Validate early: Check parameters before execution
4. Save successful patterns: Build a library

Voice Command Tips

• Be specific: Include key parameter values
• Use natural language: "Create an issue about X"
• Confirm critical: Review before destructive operations
• Learn shortcuts: Common phrases for tools

Parameter Management

• Required first: Fill mandatory fields immediately
• Defaults are smart: Often pre-filled correctly
• Save combinations: Template frequent uses
• Document patterns: Note successful parameter sets

Integration Examples

Development Workflow

Voice: "Create a bug report for the login timeout issue"

1. Tool: GitHub Create Issue
2. Auto-fills:
   - Repo: (current project)
   - Labels: ["bug"]
   - Title: "Login timeout issue"
3. You add:
   - Description details
   - Steps to reproduce
4. Execute and link to Sentry error

Monitoring Response

Voice: "Check for critical errors and create issues"

1. Tool: Sentry Get Errors
2. Filters: Severity = Critical
3. Executes and finds 3 errors
4. Follow-up: Create GitHub issues
5. Auto-populates issue details from errors

Content Publishing

Voice: "Post the weekly update to the dev community"

1. Tool: Reddit Create Post  
2. Template: "Weekly Update"
3. You modify:
   - This week's highlights
   - Links to PRs
4. Execute and share link

Troubleshooting

Parameter Issues

• Can't find option: Refresh server connection
• Validation failing: Check format requirements
• Missing fields: Ensure server is updated

Execution Problems

• Timeouts: Check network and server status
• Partial results: May need pagination
• Unexpected errors: Review server logs

Performance Tips

• Preload data: Cache common selections
• Minimize parameters: Only required fields
• Use shortcuts: Voice commands for common values
• Monitor usage: Track API limits

Master Tool Execution to transform voice commands into powerful actions!