admica/FileScopeMCP
Free分析代码库依赖关系，生成图表，帮助 AI 理解项目结构。
FreeFree tier
About admica/FileScopeMCP
FileScopeMCP is an open-source tool that analyzes codebases by identifying important files based on dependency relationships. It generates importance scores (0-10) for every file, maps bidirectional dependencies, and extracts symbols (functions, classes, interfaces, etc.) using AST-level parsing (tree-sitter) for TypeScript, JavaScript, Python, C, C++, and Rust, with regex-based support for Go, Ruby, Lua, Zig, PHP, C#, and Java. The tool features a file watcher with semantic change detection for automatic metadata updates, an LLM broker that coordinates AI work via llama.cpp or OpenAI-compatible APIs, and a Nexus web dashboard (localhost:1234) for interactive dependency graphs and per-repo health monitoring. Works with Claude Code, Hermes Agent, Codex, OpenClaw, Cursor AI, or as a standalone daemon.
Key Features
Importance ranking: every file scored 0-10 based on dependencies, exports, and location; LLM sees critical files first
Dependency mapping: bidirectional import tracking across all supported languages; detects circular dependencies
AST-level extraction (tree-sitter) for TypeScript, JavaScript, Python, C, C++, and Rust; regex-based for Go, Ruby, Lua, Zig, PHP, C#, Java
Symbol intelligence: extracts functions, classes, interfaces, types, enums, consts, modules, structs; find_symbol resolves names to file + line range
Call graph mapping: find_callers and find_callees for TypeScript/JavaScript
Always fresh: file watcher + semantic change detection; AST-level diffing for TS/JS; LLM-powered analysis for other languages; only re-processes changed files
LLM broker: background process coordinates AI work via llama.cpp's llama-server or any OpenAI-compatible HTTP API; priority queue for interactive queries
Nexus dashboard: web UI at localhost:1234 with interactive dependency graphs, file detail panels, live broker activity, per-repo health monitoring
Pros & Cons
Pros
- Open source and completely free to use
- Supports 12+ programming languages with AST-level parsing for core languages
- Automatically updates codebase metadata via file watcher and semantic change detection
- Works with multiple AI coding assistants (Claude Code, Hermes Agent, Codex, OpenClaw, Cursor AI)
- Includes a web dashboard for visual exploration of dependencies and file importance
- LLM broker handles AI tasks with priority queuing for interactive responsiveness
- Finds circular dependencies and provides call graph analysis for TypeScript/JavaScript
Cons
- Requires Node.js 22 and native build tools (may be complex for some environments)
- AST-level symbol extraction limited to TypeScript, JavaScript, Python, Go, Ruby; other languages use regex which may be less accurate
- LLM broker requires either llama.cpp or an OpenAI-compatible API, adding external dependency for full AI features
- Setup involves cloning repository and running build scripts, not a simple package install
Best For
Help AI assistants understand a codebase without reading every source fileIdentify critical files before refactoring or making changesMap dependencies and detect circular imports across multi-language projectsProvide real-time codebase summaries that stay fresh with file changesEnable AI tools like Claude Code, Cursor AI, or Codex to answer questions about specific files or functions
FAQ
What languages does FileScopeMCP support?
It supports TypeScript, JavaScript, Python, C, C++, Rust, Go, Ruby, Lua, Zig, PHP, C#, and Java. AST-level extraction (tree-sitter) is used for TS, JS, Python, C, C++, and Rust; regex-based parsing is used for the others.
How does FileScopeMCP rank file importance?
Every file is scored 0-10 based on how many things depend on it, what it exports, and where it lives. This helps AI assistants see critical files first.
Can FileScopeMCP detect circular dependencies?
Yes, it performs bidirectional import tracking and can identify circular dependencies across all supported languages.
Does it work with Claude Code or Cursor AI?
Yes, FileScopeMCP works with Claude Code, Hermes Agent, Codex, OpenClaw, Cursor AI, or can run as a standalone daemon.
Is there a way to visually explore my codebase?
Yes, the Nexus dashboard provides a web UI at localhost:1234 with interactive dependency graphs, file detail panels, live broker activity, and per-repo health monitoring.