CodeInteliMCP - Intelligent Code Analysis Server

What is CodeInteliMCP?

CodeInteliMCP is an advanced MCP (Model Context Protocol) server that combines Tree-sitter parsing with graph databases and vector search for lightning-fast code intelligence across multiple repositories. It is designed for developers who need to analyze, search, and understand codebases through Claude or other MCP-compatible clients.

How to use CodeInteliMCP?

Clone the repository and run the automated setup script (python setup_for_claude.py), or install dependencies manually with pip and add the server configuration to your Claude Desktop config file. Use the provided MCP tools to add repositories, index code, and perform semantic or graph-based searches directly from Claude.

Key features of CodeInteliMCP

• Multi-repository support for managing several codebases simultaneously
• Instant usage search to find where symbols are used in milliseconds
• Semantic code search using vector embeddings
• Dependency analysis and impact analysis for code changes
• Auto-indexing with smart incremental updates
• Multi-language support including Python, JavaScript, TypeScript, Go, Rust

Use cases of CodeInteliMCP

• Finding all usages of a function or class across a large project
• Searching for code patterns using natural language queries
• Visualizing import dependencies and class inheritance trees
• Assessing the impact of changing a symbol before making edits
• Detecting duplicated or similar code across repositories

FAQ from CodeInteliMCP

What programming languages does CodeInteliMCP support?

It supports Python, JavaScript, TypeScript, Go, Rust, and more via Tree-sitter parsers.

How is code data stored and queried?

Graph relationships (e.g., function calls, imports) are stored in DuckDB, while semantic vectors are stored in ChromaDB.

What are the runtime dependencies?

Required Python packages include mcp, duckdb, chromadb, sentence-transformers, aiofiles, and tree-sitter with language-specific grammars.

How does indexing work?

You can index an entire repository or individual files with tools like index_repository and index_file. Auto-indexing provides smart incremental updates as code changes.

What are the performance targets?

Initial indexing is under 1 minute for 100k lines of code, incremental updates under 1 second per file, usage queries under 50ms, and semantic search under 50ms.