Mission Control MCP Server

How to use

You will connect a Claude Code client to Mission Control MCP Server to define a task, run automated attempts, and receive guided feedback for each iteration. Start by registering one or more MCP servers, then define missions with a chosen validation strategy (NUMERIC, EXIT_CODE, or KEYWORD). As tasks run, submit each attempt for validation and review the feedback to refine the result until you meet the success criteria or reach the maximum number of attempts.

How to install

Prerequisites: Node.js 18 or newer, npm or yarn, and Claude Code CLI.

Option 1: NPM installation (recommended) install globally via npm and register with Claude Code.

npm install -g mission-control-mcp
```

```bash
claude mcp add mission-control mission-control-mcp
```

```
# Configuration example for Claude Code
# 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) configure CLAUDE to use a transient MCP runner.

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

Option 3: From Source (development) build and run locally.

```
# 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 for the development setup:

```
~/.config/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 shows the server connected with Claude Code.

Configuration and server endpoints

Mission Control MCP Server can be used through multiple execution methods. The following configurations are explicitly provided for running the MCP locally.

{
  "mcpServers": {
    "mission-control": {
      "type": "stdio",
      "name": "mission_control",
      "command": "mission-control-mcp",
      "args": [],
      "env": {
        "MCP_ENABLE_PERSISTENCE": "true"
      }
    }
  }
}
````

````json
{
  "mcpServers": {
    "mission-control": {
      "type": "stdio",
      "name": "mission_control_npx",
      "command": "npx",
      "args": ["-y", "mission-control-mcp"],
      "env": {
        "MCP_ENABLE_PERSISTENCE": "true"
      }
    }
  }
}
````

````json
{
  "mcpServers": {
    "mission-control": {
      "type": "stdio",
      "name": "mission_control_build",
      "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"
      }
    }
  }
}
````

You can pick any of these methods based on your preference for running the MCP locally or as part of your development workflow.

Privacy, security, and limitations

All validations run locally with zero external dependencies. Your data never leaves your machine, and there is no network latency for validations. You retain complete audit logs of every attempt and can configure resource limits to prevent excessive usage.

The MCP server is designed for objective, repeatable tasks that can be verified locally. It is not intended for subjective judgments or real-time interactive processes.

Tools and endpoints

The server exposes a set of MCP tools to manage missions and interactions. These are designed to be invoked by your Claude Code client to define missions, submit attempts, query mission status, and abort missions.

Examples and use cases

Example workflows include fixing code quality with an EXIT_CODE strategy, achieving numeric targets like test coverage or performance thresholds, validating keyword presence in outputs, and maintaining an auditable history of attempts for analysis.

define_mission

Create a new automation task with validation criteria and retry policy.

submit_attempt

Submit a task execution result for validation and receive feedback for next steps.

get_mission_status

Query the current progress and history of a mission.

abort_mission

Manually stop a running mission with an optional reason.