code-analyze-mcp

What is code-analyze-mcp?

code-analyze-mcp is a standalone Model Context Protocol server that analyzes code structure using tree-sitter. It provides AI agents with directory structure, symbol definitions, and call graphs without reading raw files, supporting 18 programming languages.

How to use code-analyze-mcp?

Install via Homebrew (brew install clouatre-labs/tap/aptu-coder), cargo-binstall, or cargo install aptu-coder. Configure your MCP client with either streamable HTTP (recommended for multi-agent setups) or stdio transport. The server exposes tools such as analyze_directory, analyze_file, analyze_symbol, edit_overwrite, edit_replace, and exec_command.

Key features of code-analyze-mcp

• Offloads mechanical code analysis, reducing token usage and costs
• Supports 18 languages including Python, Rust, TypeScript, and C++
• Provides paginated outputs for large codebases
• Includes built-in output filters for common CLI tools
• OpenSSF Silver certified for security best practices
• Offers two transport modes: streamable HTTP and stdio

Use cases of code-analyze-mcp

• AI agents exploring unfamiliar codebases via directory trees and call graphs
• Reducing token consumption during large refactoring or auditing tasks
• Tracing symbol dependencies across multi-language projects
• Replacing raw file scanning with structured structural context
• Performing targeted file edits and shell command execution within MCP

FAQ from code-analyze-mcp

What languages does code-analyze-mcp support?

It supports 18 languages: Astro, C/C++, C#, CSS, Fortran, Go, HTML, Java, JavaScript, JSON, Kotlin, Markdown, Python, Rust, TOML, TSX, TypeScript, and YAML.

How do I install code-analyze-mcp?

Install via Homebrew (brew install clouatre-labs/tap/aptu-coder), cargo binstall aptu-coder (no Rust toolchain required), or cargo install aptu-coder (requires Rust).

How do I configure code-analyze-mcp with an MCP client?

Use streamable HTTP for multi-agent setups (e.g., brew services start aptu-coder then add streamable_http configuration) or stdio for single-client use (e.g., claude mcp add --transport stdio aptu-coder -- aptu-coder).

Does code-analyze-mcp provide benchmarks showing cost savings?

Yes. Benchmarks against Django (Python) and OpenFAST (Fortran) show 14%–68% fewer tokens and 21%–68% cheaper runs compared to native tool workflows.

What runtime dependencies are required?

The server is a standalone binary; no external runtime is needed beyond the operating system. For cargo install, the Rust toolchain is required; Homebrew and cargo-binstall do not require it.