Napi is an MCP server tool from Nanoapi IO, designed for codebase intelligence and software architecture in the AI age. Built with TypeScript and running on Node.js, it functions as a local-first analysis tool. Napi helps developers, architects, and CTOs understand software structure. It supports integrating with external AI models for advanced labeling, while maintaining core functionality offline. This free tool provides machine-readable code maps essential for modern AI-driven development practices.

What does Napi connect to?

Napi connects directly to the local codebase for analysis, operating as a fully offline tool for its core functions. Optionally, it connects to external AI models such as OpenAI, Google, or Anthropic for AI-powered dependency labeling. As an MCP (Model Context Protocol) server, Napi supports HTTP as its transport protocol, using JSON-RPC 2.0 messages. It integrates with general MCP clients like Claude and ChatGPT, enabling these LLM applications to access external data sources and tools. The napi view command also starts a local web server for its interactive UI.

Core features and tools

Napi provides several callable tools for codebase management. napi init initializes a project by creating a .napirc configuration file. napi generate analyzes the codebase and produces a dependency manifest. napi view starts a local web server, opening an interactive dependency visualizer in a browser. napi extract isolates specific symbols from the codebase into separate files. Finally, napi set apiKey configures API keys for AI-powered dependency labeling, storing them locally for privacy.

Who is it for?

Napi targets developers, software architects, and CTOs who need private, machine-readable code maps for AI-assisted workflows. It is particularly useful for those prioritizing privacy in their codebase analysis and architectural understanding. The tool supports understanding dependencies, managing complexity, and tracking architectural changes. It helps professionals maintain an accurate view of their software structure. Skip this if your primary need is not local-first codebase intelligence or integration with AI models via MCP.

How to get started

No credentials are required for Napi's core functionality, as it operates as a fully offline tool. For optional AI-powered dependency labeling, API keys for providers like Google, OpenAI, or Anthropic are configured using "napi set apiKey" and stored locally in a global configuration file.

Use cases

• Codebase Analysis and Visualization: Analyze software architecture, generate dependency manifests, and visualize them through an interactive local graph to understand dependencies, complexity, and structure.
• Refactoring and Symbol Extraction: Identify and extract specific functions, classes, or symbols into separate files to support refactoring efforts.
• CI/CD Integration and Architectural Auditing: Integrate into CI/CD pipelines to automatically generate manifests, track architectural changes, and audit for files or symbols exceeding predefined complexity or coupling thresholds.

Napi offers a valuable local-first solution for codebase intelligence, supporting architectural understanding and AI integration. Its ability to generate dependency manifests and visualize them locally makes it a strong choice for privacy-conscious development teams. The tool's MCP server capabilities extend its utility by connecting LLM applications to code data.