mcp-grep-code_skill

Overview

This skill searches real-world code examples across over a million public GitHub repositories using a fast, literal-pattern code search backend. It is optimized for finding exact code snippets, usage patterns, and library calls as they appear in source files. Use it to quickly surface concrete implementations rather than topical documentation or conceptual guides.

How this skill works

The skill sends literal code-pattern queries to a remote MCP grep.app service and returns matching file snippets and repository references. Queries must be valid code fragments or regular expressions that match file contents; keyword-style searches are not effective. The tool exposes a single action, searchGitHub, which accepts a query string and returns matched code results.

When to use it

• Find concrete examples of library usage or API calls in real projects
• Locate idiomatic implementations of a function or pattern
• Audit how a function or symbol is used across many repositories
• Discover common error-handling idioms or async patterns
• Search for specific import, type, or function signatures in real code

Best practices

• Craft queries as literal code fragments or regular expressions, not plain keywords
• Inspect tool details before invoking to learn parameter schema and limits
• Keep queries short and syntactically valid so matches correspond to real files
• Use regex mode for multi-line or flexible patterns (e.g., try/await blocks)
• Adjust timeout for larger or slower searches; default timeout is 30s

Example use cases

• Find how useState is initialized across React projects with query: useState(
• Search for async/await fetch patterns using: async function fetch
• Locate library import usage with: from "openai"
• Grep for try/await structures using a dotall regex like (?s)try {.*await
• Audit how a specific helper function name is referenced across repos