Lyra Intel MCP Server

The Lyra Intel MCP (Model Context Protocol) server enables large language models like Claude to interact with Lyra Intel analysis capabilities directly.

Features

• 🔍 Codebase Analysis - Run comprehensive analysis including AST parsing, dependency mapping, and complexity metrics
• 🔎 Semantic Search - ML-powered code search to find relevant implementations
• 📊 Complexity Metrics - Get cyclomatic, cognitive, and Halstead complexity scores
• 🛡️ Security Scanning - Detect vulnerabilities, secrets, and compliance issues
• 🔬 Auto-Discovery - Scan GitHub for new MCP crypto tools and submit to registry
• ⏳ Streaming Progress - Long-running operations stream progress updates to Claude

Installation

Quick Start with Claude Code

# Install globally for one-line setup
npm install -g lyra-intel-mcp

# Add to Claude Code
claude mcp add lyra-intel -- lyra-intel-mcp

Claude Desktop

Add to your Claude Desktop config file:

macOS/Linux: ~/.config/claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "lyra-intel": {
      "command": "npx",
      "args": ["-y", "lyra-intel-mcp"]
    }
  }
}

Then restart Claude Desktop completely.

Local Development

cd mcp-server
npm install
npm run build
npm start

Usage with Claude

Natural Language

"Analyze my project at /path/to/project for security issues"
"Search for authentication patterns in the codebase"
"Check complexity of the auth module"
"Scan GitHub for new MCP crypto tools from the last 7 days"
"Analyze the repo at github.com/owner/project for MCP tools"

Tools Available

Analysis Tools

analyze-codebase

Comprehensive codebase analysis with AST parsing, dependency graphs, and metrics.

Analyze /path/to/project at deep level focusing on security and complexity

search-code

Semantic search using ML embeddings.

Search for "database connection handling" in src/

get-complexity

Get complexity metrics for code.

Show complexity metrics for src/auth/

get-security-issues

Scan for security vulnerabilities.

Find security issues in critical severity or higher

Enhanced Tools

diff-analyze

Analyze git diffs between commits/branches to understand changes, impact, and risk.

Analyze the diff between main and feature-branch in my project
Show me what changed in the last commit and its risk level

Parameters: | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | path | string | required | Path to the git repository | | sourceRef | string | HEAD~1 | Source git reference (commit, branch, tag) | | targetRef | string | HEAD | Target git reference (commit, branch, tag) | | includeImpact | boolean | true | Include impact analysis (affected tests, risk score) | | maxDepth | number | 3 | Maximum depth for impact propagation analysis | | focusFiles | string[] | - | Specific files/patterns to focus on |

Returns:

• Files changed with additions/deletions
• Semantic changes (functions, classes added/removed)
• Risk score and risk level (low/medium/high/critical)
• Impacted areas (Security, API, Data Layer, etc.)
• Actionable recommendations

generate-docs

Auto-generate documentation from code analysis.

Generate API documentation for my project in markdown format
Create a README for /path/to/codebase
Generate architecture docs for my Python project

Parameters: | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | path | string | required | Path to the codebase to document | | outputFormat | enum | markdown | Output format: markdown, html, json | | docType | enum | api | Type: api, readme, architecture, all | | includeExamples | boolean | true | Include code examples | | includeToc | boolean | true | Include table of contents | | projectName | string | - | Project name for header |

Returns:

• Generated documentation in requested format
• Sections breakdown
• Statistics (files scanned, word count)

forensic-scan

Deep analysis to find dead code, tech debt, and orphaned documentation.

Scan my project for dead code and unused functions
Find technical debt in the codebase
Run a full forensic analysis with git history

Parameters: | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | path | string | required | Path to the repository | | scanType | enum | full | Type: dead-code, tech-debt, orphans, full | | includeGitHistory | boolean | true | Include git history analysis | | minConfidence | number | 0.7 | Minimum confidence threshold (0.0-1.0) | | excludePatterns | string[] | - | Patterns to exclude from scan | | maxFileSize | number | 10 | Maximum file size in MB |

Returns:

• Dead code items with confidence scores
• Technical debt (TODOs, FIXMEs, large files, deep nesting)
• Orphaned documentation
• Git stale file analysis
• Overall health score (0-100)
• Prioritized recommendations

Discovery Tools

discovery-scan-github

Scan GitHub for new MCP crypto tool repositories.

Scan GitHub for MCP crypto repositories from the last 7 days with minimum 5 stars

Parameters: | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | daysBack | number | 7 | How many days back to search | | minStars | number | 0 | Minimum stars required | | queries | string[] | - | Custom search queries | | maxResults | number | 50 | Maximum repos to return |

discovery-analyze-repo

Analyze a GitHub repository to extract MCP tool definitions.

Analyze https://github.com/owner/mcp-crypto-tools for MCP tools

Parameters: | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | repoUrl | string | required | GitHub repository URL | | checkSecurity | boolean | true | Run security analysis |

discovery-submit-tool

Submit validated tools to the Lyra Registry.

Submit tools from github.com/owner/repo to the registry (dry run)

Parameters: | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | repoUrl | string | required | GitHub repository URL | | dryRun | boolean | true | Simulate without submitting | | minQualityScore | number | 50 | Minimum quality threshold | | minSecurityScore | number | 70 | Minimum security threshold |

discovery-run-pipeline

Run the complete discovery pipeline.

Run the full discovery pipeline for the last 3 days and submit approved tools

Parameters: | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | daysBack | number | 7 | Days to search back | | submit | boolean | false | Submit to registry | | dryRun | boolean | true | Simulate submission | | maxRepos | number | 20 | Max repos to process |

discovery-get-stats

Get discovery module statistics.

Show discovery statistics and recent runs

Integration with API

The MCP server currently streams output and basic information. For full integration with the Lyra Intel REST API:

1. Set your Lyra Intel API endpoint:

export LYRA_INTEL_API_URL=http://localhost:8000
export LYRA_INTEL_API_KEY=your_api_key

2. Tools will automatically forward requests to the backend API

Architecture

Claude/LLM Client
        ↓
    MCP Protocol
        ↓
Lyra Intel MCP Server
        ↓
Lyra Intel REST API
        ↓
Analysis Engine (Python)

Troubleshooting

"MCP server not responding"

1. Verify Node.js 16+ is installed: node --version
2. Check configuration file syntax (valid JSON)
3. Restart your MCP client completely
4. Enable debug logging:

   DEBUG=lyra-intel-mcp npm start

Large codebase timeouts

The MCP server streams progress to prevent timeouts. If you still experience timeouts:

1. Start with "quick" analysis depth
2. Use file pattern filters to narrow scope
3. Increase MCP client timeout (check client docs)

No results returned

Ensure:

1. Lyra Intel API is running and accessible
2. API credentials are set correctly
3. The provided path exists and is readable

Development

Adding New Tools

1. Create tool schema and implementation in src/tools/analysis.ts:

export const myNewTool: UnifiedTool = {
  name: "tool-name",
  description: "What it does",
  zodSchema: z.object({
    param: z.string().describe("Parameter description"),
  }),
  category: 'category',
  execute: async (args, onProgress) => {
    onProgress?.("Starting...");
    return "Result";
  }
};

2. Add to toolRegistry in src/tools/registry.ts

3. Rebuild and restart

Testing

npm run test

Documentation

For more information about Model Context Protocol, see:

• MCP Documentation
• Lyra Intel API Docs
• Lyra Intel Architecture

License

MIT - See LICENSE for details.