MCP 服务器

Code Search MCP

Enables LLMs to perform high-performance code search and analysis across multiple languages using symbol indexing, regex text search, and structural AST pattern matching. It also provides tools for technology stack detection and dependency analysis with persistent caching for optimized performance.

README

<div align="center"> <h1>Code Search MCP</h1> <p>Universal MCP server for intelligent code search across any programming language</p> </div>

<div align="center"> <h2>Overview</h2> </div>

Code Search MCP is a high-performance Model Context Protocol server that enables LLMs to intelligently search and analyze codebases across 12 programming languages with comprehensive AST search support for 15 languages. Built on our universal-ctags wrapper, ripgrep, and ast-grep, it provides fast symbol search, structural AST pattern matching, text search, file search, and dependency analysis with persistent caching for 80%+ faster startup times.

</div>

<div align="center"> <h2>Core Features</h2> </div>

<div align="center"> <table> <tr> <th>Feature</th> <th>Description</th> <th>Performance</th> </tr> <tr> <td>Symbol Search</td> <td>Find classes, functions, methods, and variables with intelligent filtering</td> <td>Fast (indexed)</td> </tr> <tr> <td>AST Search</td> <td>Structural code search using Abstract Syntax Trees with metavariables and relational rules</td> <td>Fast</td> </tr> <tr> <td>Text Search</td> <td>Regex-powered code search using ripgrep</td> <td>Very Fast</td> </tr> <tr> <td>File Search</td> <td>Locate files by name, pattern, or extension with glob support</td> <td>Very Fast</td> </tr> <tr> <td>Stack Detection</td> <td>Automatically detect technology stacks and frameworks in projects</td> <td>Fast</td> </tr> <tr> <td>Dependency Analysis</td> <td>Analyze project dependencies across multiple ecosystems</td> <td>Fast</td> </tr> <tr> <td>Index Caching</td> <td>Persistent symbol indices with automatic invalidation</td> <td>80%+ faster startup</td> </tr> </table> </div>

<div align="center"> <h2>Language Support</h2> </div>

Code Search MCP supports 12 programming languages with full symbol indexing and intelligent search capabilities.

</div>

<div align="center"> <table> <tr> <th>Language</th> <th>Symbol Search</th> <th>Text Search</th> <th>Dependency Analysis</th> </tr> <tr> <td>JavaScript</td> <td>Full</td> <td>Full</td> <td>Full (npm)</td> </tr> <tr> <td>TypeScript</td> <td>Full</td> <td>Full</td> <td>Full (npm)</td> </tr> <tr> <td>Python</td> <td>Full</td> <td>Full</td> <td>Full (pip)</td> </tr> <tr> <td>Java</td> <td>Full</td> <td>Full</td> <td>Full (Maven/Gradle)</td> </tr> <tr> <td>C#</td> <td>Full</td> <td>Full</td> <td>Full (NuGet)</td> </tr> <tr> <td>Go</td> <td>Full</td> <td>Limited</td> <td>Full (go.mod)</td> </tr> <tr> <td>Rust</td> <td>Full</td> <td>Limited</td> <td>Full (Cargo)</td> </tr> <tr> <td>C / C++</td> <td>Full</td> <td>Limited</td> <td>Limited</td> </tr> <tr> <td>PHP</td> <td>Full</td> <td>Limited</td> <td>Full (Composer)</td> </tr> <tr> <td>Ruby</td> <td>Full</td> <td>Limited</td> <td>Full (Bundler)</td> </tr> <tr> <td>Kotlin</td> <td>Full</td> <td>Limited</td> <td>Full (Gradle)</td> </tr> </table> </div>

<div align="center"> <h2>MCP Tools</h2> </div>

The server exposes the following tools through the Model Context Protocol interface.

</div>

<div align="center"> <table> <tr> <th>Tool</th> <th>Description</th> <th>Key Parameters</th> </tr> <tr> <td><code>search_symbols</code></td> <td>Search for code symbols with filters</td> <td>path, language, name, match, kinds, scope</td> </tr> <tr> <td><code>search_text</code></td> <td>Search code using regex patterns</td> <td>path, pattern, language, case_insensitive, literal, limit, paths</td> </tr> <tr> <td><code>search_files</code></td> <td>Find files by name, pattern, or extension</td> <td>path, pattern, name, extension, directory</td> </tr> <tr> <td><code>detect_stacks</code></td> <td>Detect technology stacks in a directory</td> <td>path, scan_mode (fast/thorough)</td> </tr> <tr> <td><code>analyze_dependencies</code></td> <td>Analyze project dependencies</td> <td>path, include_transitive, check_outdated</td> </tr> <tr> <td><code>refresh_index</code></td> <td>Rebuild the symbol index</td> <td>path, force_rebuild</td> </tr> <tr> <td><code>cache_stats</code></td> <td>View cache statistics</td> <td>path (optional)</td> </tr> <tr> <td><code>clear_cache</code></td> <td>Clear cached indices</td> <td>path (optional)</td> </tr> <tr> <td><code>search_ast_pattern</code></td> <td>Search using AST patterns with metavariables</td> <td>path, language, pattern, paths, limit</td> </tr> <tr> <td><code>search_ast_rule</code></td> <td>Search using complex AST rules with relational and composite operators</td> <td>path, language, rule, paths, limit, debug</td> </tr> <tr> <td><code>check_ast_grep</code></td> <td>Check ast-grep availability and version</td> <td>None</td> </tr> </table> </div>

<div align="center"> <h2>Search Capabilities</h2> </div>

<div align="center"> <table> <tr> <th>Search Type</th> <th>Match Modes</th> <th>Filter Options</th> </tr> <tr> <td>Symbol Search</td> <td>exact, prefix, substring, regex</td> <td>kind, scope (class/namespace/module), language</td> </tr> <tr> <td>Text Search</td> <td>regex, literal</td> <td>language, case sensitivity, result limit</td> </tr> <tr> <td>File Search</td> <td>glob patterns, wildcards</td> <td>extension, directory, case sensitivity</td> </tr> </table> </div>

<div align="center"> <h2>AST Search</h2> </div>

Search code using Abstract Syntax Tree analysis for structural pattern matching that goes beyond simple text search.

</div>

<div align="center"> <table> <tr> <th>Capability</th> <th>Description</th> <th>Example Pattern</th> </tr> <tr> <td>Metavariables</td> <td>Capture and match code elements</td> <td><code>$VAR</code> (named), <code>$$VAR</code> (anonymous), <code>$$$VAR</code> (multiple)</td> </tr> <tr> <td>Relational Rules</td> <td>Context-aware matching</td> <td><code>inside</code>, <code>has</code>, <code>precedes</code>, <code>follows</code></td> </tr> <tr> <td>Composite Rules</td> <td>Logical combinations</td> <td><code>all</code> (AND), <code>any</code> (OR), <code>not</code> (negation)</td> </tr> <tr> <td>Kind Matching</td> <td>Match specific AST node types</td> <td><code>function_declaration</code>, <code>class_declaration</code>, etc.</td> </tr> </table> </div>

AST Search Examples:

</div>

// Find async functions without await
{
  "rule": {
    "all": [
      { "pattern": "async function $NAME($$$) { $$$ }" },
      { "not": { "has": { "pattern": "await $$$", "stopBy": "end" } } }
    ]
  }
}

// Find React components using useEffect without dependencies
{
  "rule": {
    "all": [
      { "pattern": "useEffect($$$)" },
      { "not": { "pattern": "useEffect($CALLBACK, [$$$DEPS])" } }
    ]
  }
}

// Find functions with console.log inside
{
  "rule": {
    "pattern": "console.log($$$)",
    "inside": {
      "pattern": "function $NAME($$$) { $$$ }",
      "stopBy": "end"
    }
  }
}

Supported Languages (15 Total):

</div>

<div align="center"> <table> <tr> <th>Language</th> <th>File Extensions</th> </tr> <tr> <td>Bash</td> <td>.sh, .bash</td> </tr> <tr> <td>C</td> <td>.c, .h</td> </tr> <tr> <td>C++</td> <td>.cpp, .cc, .cxx, .hpp, .hxx</td> </tr> <tr> <td>C#</td> <td>.cs</td> </tr> <tr> <td>CSS</td> <td>.css</td> </tr> <tr> <td>Go</td> <td>.go</td> </tr> <tr> <td>HTML</td> <td>.html, .htm</td> </tr> <tr> <td>Java</td> <td>.java</td> </tr> <tr> <td>JavaScript</td> <td>.js, .jsx, .mjs</td> </tr> <tr> <td>JSON</td> <td>.json</td> </tr> <tr> <td>Kotlin</td> <td>.kt, .kts</td> </tr> <tr> <td>Python</td> <td>.py</td> </tr> <tr> <td>Rust</td> <td>.rs</td> </tr> <tr> <td>Scala</td> <td>.scala</td> </tr> <tr> <td>Swift</td> <td>.swift</td> </tr> <tr> <td>TypeScript</td> <td>.ts, .tsx</td> </tr> <tr> <td>YAML</td> <td>.yml, .yaml</td> </tr> </table> </div>

All AST language packages are bundled with the server - no additional installation required!

</div>

<div align="center"> <h2>Tech Stack Detection</h2> </div>

Automatically identify technologies, frameworks, and tools used in your projects with intelligent file-based detection.

</div>

<div align="center"> <table> <tr> <th>Category</th> <th>Technologies Detected</th> <th>Detection Method</th> </tr> <tr> <td>Languages</td> <td>JavaScript, TypeScript, Python, Java, C#, Go, Rust, C/C++, PHP, Ruby, Kotlin, Swift</td> <td>File extensions & patterns</td> </tr> <tr> <td>Build Tools</td> <td>Webpack, Vite, Rollup, Parcel, Gradle, Maven, Make, CMake, MSBuild</td> <td>Config files</td> </tr> <tr> <td>Package Managers</td> <td>npm, Yarn, pnpm, pip, Poetry, Cargo, Go modules, NuGet, Composer, Bundler</td> <td>Lock files & manifests</td> </tr> <tr> <td>Frameworks</td> <td>React, Vue, Angular, Next.js, Svelte, Django, Flask, FastAPI, Spring Boot, .NET Core</td> <td>Dependencies & configs</td> </tr> <tr> <td>Testing</td> <td>Jest, Mocha, Vitest, Pytest, JUnit, NUnit, Go Test, Cargo Test</td> <td>Config files & dependencies</td> </tr> <tr> <td>Databases</td> <td>PostgreSQL, MySQL, MongoDB, Redis, SQLite, Prisma, TypeORM, Sequelize</td> <td>Config files & dependencies</td> </tr> <tr> <td>DevOps</td> <td>Docker, Kubernetes, GitHub Actions, GitLab CI, CircleCI, Jenkins, Terraform</td> <td>Config files & manifests</td> </tr> <tr> <td>Code Quality</td> <td>ESLint, Prettier, Black, Pylint, Flake8, RuboCop, Clippy, TSLint</td> <td>Config files</td> </tr> </table> </div>

Scan Modes: Fast (config files only) • Thorough (includes dependency analysis)

</div>

<div align="center"> <h2>Performance</h2> </div>

The persistent caching system delivers dramatic performance improvements for repeated searches.

</div>

<div align="center"> <table> <tr> <th>Repository</th> <th>Cold Start</th> <th>Cached Start</th> <th>Improvement</th> </tr> <tr> <td>Express.js (8,234 symbols)</td> <td>2,453ms</td> <td>127ms</td> <td>19.3x faster</td> </tr> <tr> <td>Lodash (12,456 symbols)</td> <td>1,876ms</td> <td>89ms</td> <td>21.1x faster</td> </tr> <tr> <td>Large Codebase (5,000 symbols)</td> <td>3,124ms</td> <td>145ms</td> <td>21.5x faster</td> </tr> </table> </div>

Average improvement: 94.5% time saved

</div>

<div align="center"> <h2>Installation</h2> </div>

Prerequisites

</div>

Install the required dependencies:

# Install universal-ctags (required for symbol search)
# macOS
brew install universal-ctags

# Ubuntu/Debian
sudo apt-get install universal-ctags

# Windows (via Chocolatey)
choco install universal-ctags

# Install ripgrep (required for text search)
# macOS
brew install ripgrep

# Ubuntu/Debian
sudo apt-get install ripgrep

# Windows (via Chocolatey)
choco install ripgrep

# ast-grep is bundled with the MCP server - no separate installation needed!

Install the MCP Server

</div>

# Clone the repository
git clone https://github.com/GhostTypes/code-search-mcp.git
cd code-search-mcp

# Install dependencies
npm install

# Build the project
npm run build

<div align="center"> <h2>Configuration</h2> </div>

Security Model

All tools accept absolute path parameters and validate them against allowed workspaces configured at startup.

</div>

Add to your MCP settings file (e.g., claude_desktop_config.json):

{
  "mcpServers": {
    "code-search": {
      "command": "node",
      "args": [
        "/path/to/code-search-mcp/dist/index.js",
        "--allowed-workspace", "/path/to/your/project1",
        "--allowed-workspace", "/path/to/your/project2"
      ]
    }
  }
}

Configuration Options

</div>

<div align="center"> <table> <tr> <th>Option</th> <th>Description</th> </tr> <tr> <td><code>--allowed-workspace <path></code></td> <td>Whitelist a directory for search operations. Can be specified multiple times. If omitted, all paths are allowed (use with caution)</td> </tr> <tr> <td><code>-w <path></code></td> <td>Short alias for <code>--allowed-workspace</code></td> </tr> </table> </div>

<div align="center"> <h2>Development</h2> </div>

# Clone the repository
git clone https://github.com/GhostTypes/code-search-mcp.git
cd code-search-mcp

# Install dependencies
npm install

# Build the project
npm run build

# Run tests
npm test

# Run integration tests
npm run test:integration

<div align="center"> <h2>Architecture</h2> </div>

The server is built with a modular architecture for maintainability and extensibility.

</div>

<div align="center"> <table> <tr> <th>Component</th> <th>Responsibility</th> </tr> <tr> <td>MCP Server</td> <td>Protocol handling, tool routing, and path validation</td> </tr> <tr> <td>Symbol Indexer</td> <td>Universal-ctags integration and indexing</td> </tr> <tr> <td>Symbol Search Service</td> <td>Symbol query processing and filtering</td> </tr> <tr> <td>Text Search Service</td> <td>Ripgrep integration for text search</td> </tr> <tr> <td>File Search Service</td> <td>Fast file finding with glob patterns</td> </tr> <tr> <td>Stack Detection Engine</td> <td>Technology stack identification</td> </tr> <tr> <td>Dependency Analyzer</td> <td>Multi-ecosystem dependency analysis</td> </tr> <tr> <td>Cache Manager</td> <td>Index persistence and invalidation</td> </tr> <tr> <td>AST Search Service</td> <td>Structural code search using ast-grep</td> </tr> </table> </div>

<div align="center"> <h2>Contributing</h2> </div>

Contributions are welcome! Please feel free to submit issues or pull requests.

</div>

<div align="center"> <h2>License</h2> </div>

MIT License - see LICENSE for details

</div>

<div align="center"> <h2>Acknowledgments</h2> </div>

<div align="center"> <table> <tr> <th>Tool</th> <th>Purpose</th> </tr> <tr> <td><a href="https://ctags.io/">universal-ctags</a></td> <td>Symbol indexing</td> </tr> <tr> <td><a href="https://github.com/BurntSushi/ripgrep">ripgrep</a></td> <td>Text search</td> </tr> <tr> <td><a href="https://ast-grep.github.io/">ast-grep</a></td> <td>AST-based structural search</td> </tr> <tr> <td><a href="https://github.com/modelcontextprotocol/sdk">MCP SDK</a></td> <td>Protocol implementation</td> </tr> </table> </div>

Code Search MCP

README

推荐服务器