Mission Control MCP Server

Mission Control MCP | Mission Control Automation Orchestrator

A domain-agnostic task automation and validation system for Claude Code, enabling autonomous task execution with self-verification capabilities.

📋 Overview

Mission Control is a Model Context Protocol (MCP) server designed specifically for Claude Code. It provides a framework for automating tasks with built-in validation loops, allowing Claude to:

• Execute tasks autonomously
• Verify results against objective criteria
• Iterate until success or maximum attempts
• Provide detailed feedback for each attempt

Key Features

• ✅ Zero External Dependencies: All validation runs locally
• 🔒 Privacy-Focused: No data leaves your machine
• ⚡ Fast: No network latency for validations
• 🎯 Domain-Agnostic: Works across software development, DevOps, data science, and more
• 📊 Self-Validating: Automated success/failure detection
• 🔄 Iterative: Automatic retry with feedback

🏗️ Architecture

Core Components

1. Mission Definer: Locks in task parameters and validation criteria
2. Attempt Counter: Prevents infinite loops with configurable limits
3. Dynamic Feedback Engine: Generates actionable improvement suggestions
4. Validation Strategies: NUMERIC, EXIT_CODE, and KEYWORD validators
5. State Manager: Tracks mission progress and history

Workflow

┌─────────────────────────────────────────┐
│ 1. Define Mission                       │
│    - Goal description                   │
│    - Validation strategy                │
│    - Success criteria                   │
└─────────────────────────────────────────┘
              ↓
┌─────────────────────────────────────────┐
│ 2. Execution Loop                       │
│    ┌─────────────────────────────────┐  │
│    │ Claude executes task            │  │
│    └─────────────────────────────────┘  │
│              ↓                          │
│    ┌─────────────────────────────────┐  │
│    │ Submit attempt for validation   │  │
│    └─────────────────────────────────┘  │
│              ↓                          │
│    ┌─────────────────────────────────┐  │
│    │ System validates result         │  │
│    └─────────────────────────────────┘  │
│              ↓                          │
│        Pass? ─No→ Retry with feedback   │
│          │                              │
│         Yes                             │
└─────────────────────────────────────────┘
              ↓
┌─────────────────────────────────────────┐
│ 3. Mission Complete                     │
│    - Success report                     │
│    - Attempt history                    │
└─────────────────────────────────────────┘

📦 Installation

Prerequisites

• Node.js >= 18.0.0
• npm or yarn
• Claude Code CLI

Option 1: NPM Installation (Recommended)

Install globally via npm:

npm install -g mission-control-mcp

Then add to Claude Code:

claude mcp add mission-control mission-control-mcp

Or configure manually:

Configuration file location:

• macOS/Linux: ~/.config/claude-code/mcp.json
• Windows: %USERPROFILE%\.claude-code\mcp.json

{
  "mcpServers": {
    "mission-control": {
      "command": "mission-control-mcp",
      "env": {
        "MCP_ENABLE_PERSISTENCE": "true"
      }
    }
  }
}

Option 2: Using npx (No Installation)

claude mcp add mission-control --transport stdio -- npx -y mission-control-mcp

Option 3: From Source (Development)

# Clone the repository
git clone https://github.com/ufo23005/mission-control-mcp.git
cd mission-control-mcp

# Install dependencies
npm install

# Build the project
npm run build

Configure Claude Code:

Configuration file location:

• macOS/Linux: ~/.config/claude-code/mcp.json
• Windows: %USERPROFILE%\.claude-code\mcp.json

{
  "mcpServers": {
    "mission-control": {
      "command": "node",
      "args": ["<path-to-installation>/mission-control-mcp/dist/index.js"],
      "env": {
        "MCP_STATE_DIR": "<path-to-installation>/mission-control-mcp/.state",
        "MCP_ENABLE_PERSISTENCE": "true"
      }
    }
  }
}

Verify Installation

claude mcp list
# Should show mission-control in the list with ✓ Connected status

🚀 Usage

Validation Strategies

1. NUMERIC - Number Comparison

For tasks with measurable numeric goals (performance metrics, coverage, scores).

Example: Test Coverage

// Define mission
{
  "goal": "Increase test coverage to 80%",
  "strategy": "NUMERIC",
  "criteria": {
    "operator": ">=",
    "threshold": 80.0,
    "metricName": "coverage_percentage"
  },
  "maxAttempts": 10
}

// Submit attempt
{
  "missionId": "<mission-id>",
  "output": "Coverage: 82.5%",
  "value": 82.5
}

Operators: >, <, >=, <=, ==, !=

2. EXIT_CODE - Command Status

For tasks validated by command execution success.

Example: Linting

// Define mission
{
  "goal": "Fix all ESLint errors",
  "strategy": "EXIT_CODE",
  "criteria": {
    "expectedCode": 0,
    "command": "npm run lint"
  },
  "maxAttempts": 8
}

// Submit attempt
{
  "missionId": "<mission-id>",
  "output": "Linting complete",
  "value": 0  // exit code
}

3. KEYWORD - Text Matching

For tasks validated by output content.

Example: Build Success

// Define mission
{
  "goal": "Successful production build",
  "strategy": "KEYWORD",
  "criteria": {
    "keyword": "Build completed successfully",
    "mustContain": true,
    "caseSensitive": false
  }
}

// Submit attempt
{
  "missionId": "<mission-id>",
  "output": "Build completed successfully. Output: dist/"
}

🛠️ MCP Tools

define_mission

Define a new automation task with validation criteria.

Parameters:

• goal (string): Natural language task description
• strategy (enum): NUMERIC | EXIT_CODE | KEYWORD
• criteria (object): Strategy-specific validation parameters
• maxAttempts (number, optional): Maximum retry limit (default: 10)
• enableCheckpoints (boolean, optional): Enable recovery checkpoints
• checkpointFrequency (number, optional): Checkpoint every N attempts

Returns: Mission ID and configuration

submit_attempt

Submit a task execution result for validation.

Parameters:

• missionId (string): Mission identifier
• output (string): Command/process output
• value (number|string, optional): Extracted validation value

Returns: Validation result with feedback and next steps

get_mission_status

Query current mission progress and history.

Parameters:

• missionId (string): Mission identifier

Returns: Status, attempt count, and history

abort_mission

Manually stop a running mission.

Parameters:

• missionId (string): Mission identifier
• reason (string, optional): Abort reason

Returns: Confirmation and final status

📖 Examples

Example 1: Fix ESLint Errors

$ claude "Fix all ESLint errors. Use define_mission with EXIT_CODE strategy, then iterate until npm run lint returns 0."

Claude will:

1. Call define_mission with EXIT_CODE criteria (expectedCode: 0)
2. Analyze code and fix errors
3. Run npm run lint
4. Call submit_attempt with exit code
5. Repeat until successful or max attempts

Example 2: Optimize Performance

$ claude "Reduce API response time below 100ms. Validate with local benchmark tests."

Claude will:

1. Define mission with NUMERIC strategy (< 100)
2. Make optimizations
3. Run benchmark tests
4. Submit results
5. Iterate based on feedback

Example 3: Documentation Compliance

$ claude "Ensure all public functions have JSDoc comments. Check with custom validator."

Claude will:

1. Define mission with KEYWORD or custom validation
2. Add missing documentation
3. Run documentation checker
4. Verify completeness

🔧 Configuration

State Management

The orchestrator maintains state in .state/ directory:

.state/
├── missions/       # Active and completed missions
└── checkpoints/    # Recovery checkpoints

Retention Policies

• Completed missions: 30 days (configurable)
• Failed missions: 90 days (for analysis)
• Active missions: Retained until completion

Resource Limits

Default limits (configurable in code):

• Max attempts: 10
• Checkpoint frequency: Every 5 attempts
• Attempt timeout: None (per-mission configurable)

📁 Project Structure

src/
├── index.ts              # MCP server entry point
├── types/                # TypeScript type definitions
│   ├── mission.ts        # Mission and attempt types
│   ├── validation.ts     # Validation strategy types
│   └── state.ts          # State management types
├── core/                 # Core components
│   ├── MissionDefiner.ts # Mission configuration
│   ├── AttemptCounter.ts # Retry management
│   └── FeedbackEngine.ts # Feedback generation
├── validators/           # Validation strategies
│   ├── BaseValidator.ts
│   ├── NumericValidator.ts
│   ├── ExitCodeValidator.ts
│   └── KeywordValidator.ts
├── state/                # State management
│   └── StateManager.ts   # Mission state tracking
├── tools/                # MCP tool implementations
│   ├── defineMission.ts
│   ├── submitAttempt.ts
│   ├── getMissionStatus.ts
│   └── abortMission.ts
└── utils/                # Utilities
    ├── logger.ts         # Logging system
    └── errors.ts         # Error types

🎯 Use Cases

| Domain | Task | Strategy | Success Criteria | | ----------------- | -------------------- | --------- | -------------------------- | | Software Dev | Fix lint errors | EXIT_CODE | exit_code == 0 | | Testing | Achieve 80% coverage | NUMERIC | coverage >= 80 | | Performance | Optimize load time | NUMERIC | load_time < 2.0 | | DevOps | Verify deployment | KEYWORD | Output contains "deployed" | | Data Science | Improve model F1 | NUMERIC | f1_score >= 0.85 | | Documentation | Complete API docs | KEYWORD | No "TODO" found |

🔒 Privacy & Security

• No external APIs: All validation runs locally
• No data transmission: Everything stays on your machine
• No network required: Works completely offline
• Audit logging: Complete attempt history
• Configurable limits: Prevent resource exhaustion

🚧 Limitations

Suitable For:

• Objectively measurable outcomes
• Repeatable operations
• Automatable verification
• Local command-line tools

Not Suitable For:

• Subjective quality judgments
• One-time irreversible operations
• Tasks requiring human creativity
• Real-time interactive processes

🤝 Contributing

Contributions welcome! Please ensure:

• All TypeScript types are defined
• Code follows existing patterns
• Tests pass (when implemented)
• Documentation updated

📄 License

MIT

📚 References

• Model Context Protocol - MCP specification
• Claude Code MCP - Claude Code MCP documentation