lyra-tool-discovery

Lyra Tool Discovery is an AI-powered tool designed to automate the discovery and preparation of MCP (Model Context Protocol) plugins and servers in the crypto, DeFi, blockchain, and web3 space. It crawls sources like GitHub and npm to identify crypto-relevant MCP servers, analyzes them with AI, and outputs ready-to-use plugin configurations and manifests. The CLI offers templates, analysis commands, and JSON output suitable for CI/CD pipelines, making it ideal for plugin marketplace operators and automation engineers who need to rapidly onboard new tools.

You can install and run Lyra either globally or on-demand via npx. Use the discover workflow to search for MCP tools, analyze a specific repository or npm package, and generate plugin manifests. Key commands include discovering tools from sources, analyzing a repo or npm package, listing available AI providers, and exploring plugin templates. The tool supports dry-run mode for safe testing and provides structured JSON output for easy integration into automation pipelines.

Prerequisites:

• Node.js (>= 14) and npm or pnpm
• Optional: a GitHub token for higher API rate limits

Install options:

1. Global CLI installation (recommended for CLI usage)

   • npm: npm install -g @nirholas/lyra-tool-discovery
   • pnpm: pnpm add -g @nirholas/lyra-tool-discovery

2. Run directly with npx (no install)

   • npx @nirholas/lyra-tool-discovery discover --help

3. Project-local installation (as a project dependency)

   • npm: npm install @nirholas/lyra-tool-discovery
   • pnpm: pnpm add @nirholas/lyra-tool-discovery

Usage example:

• lyra-discover discover --sources github,npm --limit 10
• lyra-discover analyze-repo anthropics claude-mcp

Environment variables (optional but recommended):

• OPENAI_API_KEY="sk-..." or ANTHROPIC_API_KEY="sk-..."
• AI_PROVIDER="openai" or "anthropic" (defaults to auto-detect)
• AI_MODEL="claude-sonnet-4-20250514" (optional)
• GITHUB_TOKEN="ghp_..." (optional for rate limits)

Tips and common considerations:

• The Lyra tool supports multiple AI providers and models; set AI_PROVIDER and AI_MODEL to control behavior.
• For heavy usage or CI/CD pipelines, provide a GitHub token to avoid rate limits when discovering from GitHub.
• Use the dry-run option to test discovery results without invoking AI analysis.
• The mcp_config shown maps to running the Lyra CLI as a node-based MCP server via npx; adjust if you host a dedicated server or wrapper around the CLI.
• If you plan to contribute, familiarize yourself with the available templates (e.g., mcp-http, mcp-stdio, openapi, standalone) to tailor outputs to your MCP ecosystem.