SHIP Protocol MCP Server

AI coding reliability scoring for Claude Desktop and Cursor via Model Context Protocol (MCP).

What is SHIP?

SHIP (Success Heuristics for Intelligent Programming) is the industry standard for AI coding reliability. It predicts whether an AI coding task will succeed before you start.

SHIP Score is calculated from 4 layers:

• Confidence (40%): Task complexity, historical success, pattern matching
• Focus (30%): Prompt clarity, context relevance, scope definition
• Context (20%): Code context completeness, quality, freshness
• Efficiency (10%): Token usage, redundancy, cost optimization

Quick Start

Claude Desktop

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on Mac or %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "ship": {
      "command": "npx",
      "args": ["-y", "@vibeatlas/ship-mcp-server"]
    }
  }
}

Cursor

Add to your Cursor MCP settings:

{
  "mcpServers": {
    "ship": {
      "command": "npx",
      "args": ["-y", "@vibeatlas/ship-mcp-server"]
    }
  }
}

With API Key (Optional)

For personalized predictions:

{
  "mcpServers": {
    "ship": {
      "command": "npx",
      "args": ["-y", "@vibeatlas/ship-mcp-server"],
      "env": {
        "SHIP_API_KEY": "your-api-key-here"
      }
    }
  }
}

Available Tools

1. ship_assess - Assess Task Reliability

Assess an AI coding task before starting.

Example:

"Assess this task: Implement user authentication with JWT in TypeScript"

Returns:

• SHIP Score (0-100) with grade (A+ to F)
• Layer-by-layer breakdown
• Specific recommendations for improvement

2. ship_feedback - Submit Outcome

Report the outcome of a task to improve predictions.

Example:

"Submit feedback for request abc123: outcome was success with 1 attempt"

3. ship_score - Quick Score

Get an instant reliability score without full assessment.

Example:

"Get quick SHIP score: complexity 5, context relevance 0.8"

4. ship_metrics - View Statistics

View aggregate metrics and patterns.

Example:

"Show SHIP metrics for the past week"

5. ship_health - Check API Status

Verify the SHIP API is available.

Example:

"Check SHIP API health"

6. ship_badge - Get Project Badge

Get a badge URL for your README.

Example:

"Get SHIP badge for project my-awesome-app"

7. ship_interpret - Interpret Score

Get human-readable interpretation of a score.

Example:

"Interpret SHIP score 75"

Usage Examples

Before Starting a Complex Task

Human: I want to implement a real-time chat feature with WebSocket support.

Claude: Let me assess this task first with SHIP...

[Uses ship_assess tool]

The SHIP Score is 72 (Grade B). The assessment shows:
- Confidence: 68/100 - This is a moderately complex task
- Focus: 85/100 - Your request is clear
- Context: 60/100 - Would benefit from more code context
- Efficiency: 78/100 - Good token efficiency

Recommendations:
- Provide existing server configuration
- Share any current WebSocket implementations
- Clarify real-time requirements (latency, scale)

Would you like to proceed or provide more context first?

After Completing a Task

Human: That worked perfectly, thanks!

Claude: Great! Let me submit feedback to improve SHIP accuracy.

[Uses ship_feedback tool with outcome: success]

Feedback submitted. This helps SHIP make better predictions for similar tasks.

API Reference

SHIP API Base URL

https://ship.vibeatlas.dev

(Fallback: https://ship-protocol.dhruvaapi.workers.dev)

Endpoints

| Endpoint | Method | Description | |----------|--------|-------------| | /v1/health | GET | API health check | | /v1/ship/assess | POST | Full task assessment | | /v1/ship/feedback | POST | Submit task outcome | | /v1/ship/score | POST | Quick score calculation | | /v1/ship/metrics | GET | Aggregate metrics | | /v1/ship/badge/:id | GET | Badge SVG/data |

Grade Scale

| Grade | Score Range | Meaning | |-------|-------------|---------| | A+ | 95-100 | Exceptional - Will almost certainly succeed | | A | 85-94 | High reliability - Very likely to succeed | | B | 70-84 | Good - Should work with minor issues | | C | 50-69 | Moderate - May need iteration | | D | 30-49 | Low - High risk of failure | | F | 0-29 | Unreliable - Likely to fail |

Development

Build from Source

cd packages/ship-mcp-server
npm install
npm run build

Run Locally

npm run dev

Test

npm test

Environment Variables

| Variable | Description | Default | |----------|-------------|---------| | SHIP_API_KEY | API key for personalized predictions | (optional) | | SHIP_API_URL | Custom API URL | https://ship-protocol.dhruvaapi.workers.dev |

Related

• SHIP Protocol Specification
• VibeAtlas - AI Context Intelligence Platform
• @vibeatlas/ship-protocol - SHIP Protocol SDK

License

MIT - See LICENSE

SHIP Protocol - The trust layer for AI coding

Made with ❤️ by VibeAtlas