codeweave

codeweave

A self-hosted MCP server that indexes your codebase and provides AI assistants with deep context including file tree, full-text search, git history, dependencies, and stack detection, all without sending your code to third parties.

Category
访问服务器

README

codeweave

npm version CI License: MIT npm downloads

Make any repo instantly AI-ready via MCP — a self-hosted, zero-API-key context server for your codebase.

<div align="center"> <img src="assets/demo.gif" alt="codeweave demo" width="800" /> </div>

codeweave runs locally alongside your AI coding tool. It indexes your entire repo into a fast SQLite database and exposes it as an MCP server, giving AI assistants deep, accurate context about your code — file tree, full-text search, git history, dependencies, stack detection, and coding conventions — without ever sending your code to a third party.


Quick Start

# 1. Install globally (or use npx)
npm install -g codeweave

# 2. Generate a config in your project root
cd /path/to/your/project
codeweave init

# 3. Start the MCP server
codeweave start

Then add codeweave to your AI client (see AI Client Setup below).


Features

Feature Description
File tree Annotated JSON tree with language tags for every file
Full-text search SQLite FTS5 with porter stemming — sub-millisecond queries
File content Line-numbered file content with optional line range slicing
Git history Recent commits with author, date, and changed files
Dependency graph Parsed from package.json, Cargo.toml, go.mod, pyproject.toml, and 6 more
Stack detection Automatically detects language, framework, database, test runner, styling
Convention detection File naming style, import patterns, component structure
File watcher Re-indexes changed files in < 1s with 300ms debounce
Incremental indexing mtime-based cache — warm starts in < 30ms
Zero API keys Runs entirely locally. No cloud. No telemetry.

AI Client Setup

codeweave uses the stdio MCP transport and works with any MCP-compatible client.

Claude Code

Add to .claude/settings.json in your project root:

{
  "mcpServers": {
    "codeweave": {
      "command": "npx",
      "args": ["codeweave"],
      "cwd": "."
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "codeweave": {
      "command": "npx",
      "args": ["codeweave"],
      "cwd": "${workspaceFolder}"
    }
  }
}

Zed

Add to ~/.config/zed/settings.json:

{
  "context_servers": {
    "codeweave": {
      "command": {
        "path": "npx",
        "args": ["codeweave"]
      }
    }
  }
}

Continue.dev

Add to .continue/config.json:

{
  "mcpServers": [
    {
      "name": "codeweave",
      "command": "npx",
      "args": ["codeweave"]
    }
  ]
}

CLI Reference

codeweave [command] [options]

Commands:
  start    Start the MCP server (default)
  init     Generate codeweave.config.js in the current directory
  status   Show indexing stats for the current repo

Options:
  --verbose    Enable debug logging
  -V, --version  Show version number
  -h, --help     Show help

See the full CLI reference →


Configuration

Run codeweave init to generate a codeweave.config.js. All fields are optional — defaults work well for most projects.

// codeweave.config.js
export default {
  // Directories to index (relative to config file)
  include: ['src'],

  // Extra patterns to exclude beyond .gitignore
  exclude: ['**/*.generated.ts'],

  // Files larger than this are truncated (KB)
  maxFileSizeKB: 100,

  // How many git commits to read
  gitDepth: 50,

  // Port for future HTTP transport
  port: 3333,

  // Enable verbose/debug logging
  verbose: false,
};

See the full configuration reference →


MCP Tools

codeweave exposes 7 MCP tools that AI assistants can call:

Tool Description
get_file_tree Annotated file tree, filterable by language
get_file File content with line numbers and optional line range
search_codebase Full-text search with snippet context
get_conventions Detected naming style, import patterns, test framework
get_dependencies All dependencies from parsed manifest files
get_git_history Recent commits, filterable by file path
get_stack_info Detected language, framework, database, tooling

See the full API reference →


Architecture

your repo on disk
      │
      ▼
┌─────────────────────────────────────────────────────┐
│                   codeweave process                  │
│                                                     │
│  ┌──────────┐  ┌──────────┐  ┌───────────────────┐ │
│  │ File     │  │   Git    │  │  Manifest Parser  │ │
│  │ Scanner  │  │  Reader  │  │  (9 formats)      │ │
│  └────┬─────┘  └────┬─────┘  └────────┬──────────┘ │
│       │              │                 │            │
│       └──────────────┴─────────────────┘            │
│                       │                             │
│              ┌────────▼────────┐                   │
│              │  SQLite DB      │                   │
│              │  (WAL + FTS5)   │◄──────────────────┤
│              └────────┬────────┘     File Watcher  │
│                       │                             │
│       ┌───────────────┼───────────────┐            │
│       ▼               ▼               ▼            │
│  get_file_tree   search_codebase  get_stack_info   │
│  get_file        get_conventions  get_dependencies │
│  get_git_history                                   │
│       │                                            │
│       └──────────────────────────────────────────┐ │
│                    MCP Server (stdio)             │ │
└───────────────────────────────────────────────────┘
                        │
              ┌─────────┘
              ▼
    AI client (Claude Code, Cursor, Zed…)

Key design decisions:

  • SQLite over in-memory — the full file content lives on disk, not in the Node.js heap. Memory usage stays flat regardless of repo size.
  • FTS5 porter tokenizer — full-text search that handles stemming (authenticateauthenticat) without any external dependencies.
  • mtime cache — unchanged files are never re-read. Warm starts on a 1K-file repo take ~28ms.
  • Stdio transport — no port conflicts, no firewall rules. Works in any environment where the AI client can spawn a subprocess.

Supported Stacks

codeweave detects and parses:

Languages: TypeScript, JavaScript, Python, Go, Rust, Dart, Ruby, PHP, Java, Kotlin, Swift, and more

Frameworks: Next.js, Nuxt, Remix, SvelteKit, Astro, Vite+React, Express, Fastify, NestJS, Hono, Django, Flask, FastAPI, Rails, Laravel, Symfony, Angular, Vue, Svelte

Databases: Prisma, TypeORM, Drizzle, MongoDB, PostgreSQL, MySQL, SQLite, Redis, Sequelize

Manifest formats: package.json, pubspec.yaml, Cargo.toml, go.mod, requirements.txt, pyproject.toml, composer.json, Gemfile, build.gradle


Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines on how to fork, develop, test, and submit PRs.

git clone https://github.com/MdAbdullahAlMahmud/codeweave.git
cd codeweave
npm install
npm test         # run tests
npm run typecheck  # type check
npm run lint     # lint

License

MIT © Abdullah — see LICENSE for details.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选