nobiehl/codeweaver-mcp

CodeWeaver 🕸️

⚠️ Beta Release (v0.6.0) - Static Analysis + Code Cleanup! SpotBugs & Checkstyle integration, all packages updated, codebase fully cleaned. PMD & SonarLint coming soon!

Token-efficient MCP server for Multi-Language project analysis (Java, TypeScript, JavaScript, Markdown, Python)

Weaving Code Intelligence for LLMs - A lightweight Model Context Protocol server that provides token-efficient access to Java, TypeScript, JavaScript, Markdown, and Python files through a multi-agent architecture with language plugins.

⚡ Highlights

• ✅ Zero Native Dependencies* - Pure Node.js/TypeScript with language-specific parsers
• ✅ Dual Interface - CLI tool AND MCP server from same codebase
• ✅ Token-Efficient - Smart file reading with token limits
• ✅ Multi-Language Support 🆕 - Java, TypeScript, JavaScript, Markdown, AND Python with unified plugin architecture!
  • 🎯 Complete Java Support (Java 8-23) - Records, Sealed Classes, Module System
  • 🎯 Complete TypeScript Support - Classes, Interfaces, Types, Enums, Generics, Decorators
  • 🎯 Complete JavaScript Support - Modern ES6+, JSX, Arrow Functions, Async/Await
  • 🎯 Complete Markdown Support - Headers as Sections, Links as References, Code Blocks
  • 🎯 Complete Python Support 🆕 - Classes, Functions, Methods, Decorators, Type Hints, Async/Await
  • ✅ Class-Level Annotations/Decorators - Spring, JPA, Jakarta EE, TypeScript decorators, Python @decorators
  • ✅ Method Parameters - Names, types, and annotations extracted
  • ✅ Generic Type Parameters - Full signature with bounds (Java, TypeScript)
  • ✅ Language Field - Every symbol tagged with source language
  • ✅ Easy Extensibility - Plugin system for adding new languages
• ✅ Powerful Search - Keyword, pattern, AND semantic search (AI-powered) 🆕
• ✅ Semantic Code Search - Find code by meaning/intent using LanceDB + Transformers 🆕
  • ⚡ ONNX Runtime Optimizations - Multi-threading + SIMD for 3x faster embeddings! 🆕
  • ⚡ 16x faster with Batch-Processing - 10k files in ~10 min (was 8h!)
  • 🎯 Multi-Collection Support - Separate indexes for Code AND Docs! 🆕
  • 🔍 File Watcher - Automatic incremental updates on file changes! 🆕
  • 📖 - Comprehensive guide with workflows and best practices
  • 🎯 - Multi-collection usage guide (Code + Docs)
  • 🔍 - Keep your index always up-to-date!
  • 🚀 - Future optimizations (GPU acceleration)
• ✅ Code Quality Analysis - Cyclomatic complexity, LOC metrics, import analysis
• ✅ Static Analysis 🆕 - SpotBugs (bugs), Checkstyle (style) with plugin architecture!
  • 🔬 SpotBugs - Finds NullPointerExceptions, Resource Leaks, SQL Injections
  • ✅ Checkstyle - Code style enforcement, naming conventions
  • 🔌 Plugin Architecture - Easy to add PMD, SonarLint later
• ✅ Git Integration - Status, diff, blame, log, branches, compare
• ✅ Test-Driven - 291 tests passing (100%) - All features fully tested! 🆕

* Core features (Discovery, Symbols, Search, Analysis, VCS) have zero native dependencies. Semantic Search optionally requires LanceDB + ONNX Runtime (native components).

🚀 Quick Start

# Install
npm install

# Build
npm run build

# Use as CLI
npm run dev -- info
npm run dev -- symbols index
npm run dev -- search keyword "CodeWeaver"
npm run dev -- analysis project

# Use as MCP Server
npm run dev -- --mcp

📖 Documentation

📚 - Full navigation of all CodeWeaver documentation

Quick Links:

New to CodeWeaver? Start with (5 minutes) or (20 minutes).

📦 Current Features (Phase 1 + 2 + 3 + 4 Complete)

✅ Implemented

Agents:

• Project Metadata Agent 🆕 - Multi-language project metadata extraction with plugin architecture
  • Gradle: Java version, dependencies, plugins, modules
  • npm: TypeScript/JavaScript, package manager detection, scripts, workspaces
  • Auto-detection: Automatically detects project type(s)
  • Unified Schema: Language-agnostic metadata format
  • Extensible: Easy to add new build systems (pip, Maven, Cargo, etc.)
• System Check Agent 🆕 - Dependency validation and system health checks
  • Critical: Node.js (>=18), Git (>=2.0) - required for core features
  • Optional: Python (>=3.8), Gradle (>=7.0), Maven (>=3.6)
  • Auto-check: Quick startup validation in CLI mode
  • Doctor Command: Full system diagnostic with recommendations
• Cache Agent - Content-addressable caching with SHA-256 hashing
• Snippets Agent - Token-efficient file reading with line ranges
• Symbols Agent - Multi-language symbol extraction with plugin architecture 🆕
  • Java: Classes, Interfaces, Enums, Records, Annotation Types, Sealed Classes, Module System
  • TypeScript: Classes, Interfaces, Types, Enums, Functions, Generics, Decorators, Namespaces
  • JavaScript: Classes, Functions, Arrow Functions, Async/Await, ES6+ features
  • Markdown: Headers as Sections, Local Links as References, Code Blocks
  • Python 🆕: Classes, Functions, Methods, Decorators, Type Hints, Async/Await
  • Methods with parameters, generics, and annotations/decorators
  • Fields/Properties with modifiers and visibility
  • Constructors, nested types, enum constants
  • Language-tagged symbols for easy filtering
• Search Agent - Keyword and pattern search with file filtering
• Analysis Agent - Cyclomatic complexity, LOC metrics, code quality
• VCS Agent - Git operations (status, diff, blame, log, branches, compare)
• Semantic Index Agent - LanceDB vector search with multi-collection support 🆕
• File Watcher Agent - Automatic incremental index updates on file changes 🆕
• Symbol Storage - In-memory symbol index with JSON Lines persistence

MCP Tools (19 total):

File & Project:

• project.meta - Get unified project metadata (auto-detects: Gradle, npm, pip, Maven, etc.)
  • Multi-language support with plugin architecture
  • Optional projectType parameter for specific extraction
• file.read - Read file with optional token limit (default: 10000)
• file.readRange - Read specific line ranges (1-indexed, inclusive)
• file.readWithNumbers - Read file with line numbers for reference

Symbols:

• symbols.index - Index entire project and extract symbols
• symbols.find - Find symbols by name (case-insensitive substring)
• symbols.findByKind - Find symbols by kind (class/method/field/constructor)
• symbols.get - Get symbol details by qualified name

Search:

• search.keyword - Search for keyword in files (grep-like)
• search.files - Find files by name pattern (glob-like: *.java)

Analysis:

• analysis.file - Analyze single file for complexity and metrics
• analysis.project - Analyze entire project for statistics

Version Control:

• vcs.status - Get Git repository status
• vcs.diff - Get diff for file(s)
• vcs.blame - Get Git blame information for file
• vcs.log - Get commit history
• vcs.branches - Get list of all branches
• vcs.compare - Compare two branches

System:

• system.check - Check system dependencies (planned for MCP integration)

CLI Commands:

Info & Files:

• codeweaver info - Display project information
• codeweaver file read <path> [--limit N] [--numbers] - Read files
• codeweaver file range <path> <start> <end> - Read line ranges
• codeweaver file context <path> <line> [-c N] - Get context around line

Symbols:

• codeweaver symbols index - Index project and extract symbols
• codeweaver symbols find <name> - Find symbols by name
• codeweaver symbols get <qualifiedName> - Get symbol details
• codeweaver symbols list <kind> - List all symbols of a kind

Search:

• codeweaver search keyword <keyword> [-i] [-m N] [-c N] [-e .ext] - Keyword search
• codeweaver search files <pattern> - Find files by pattern
• codeweaver search semantic <query> [--index] [-c collection] [-l N] - Semantic search 🆕

Analysis:

• codeweaver analysis file <path> - Analyze file complexity and metrics
• codeweaver analysis project [--top N] - Analyze project statistics
• codeweaver analysis complexity <path> - Show complexity breakdown

Version Control:

• codeweaver vcs status - Show Git repository status
• codeweaver vcs diff [file] - Show diff for file(s)
• codeweaver vcs blame <file> [-l <range>] - Show Git blame
• codeweaver vcs log [-n N] [--since] [--author] - Show commit history
• codeweaver vcs branches - List all branches
• codeweaver vcs compare <base> <compare> - Compare two branches

File Watching: 🆕

• codeweaver watch [--debounce N] [--code-only] [--docs-only] - Watch files and auto-update index

System Check: 🆕

• codeweaver doctor - Check system dependencies (Node.js, Git, Python, Gradle, Maven)
• codeweaver doctor --quick - Quick check (only critical dependencies)

Infrastructure:

• Auto-detection (stdio = MCP mode, TTY = CLI mode)
• Progress tracking (JSON Lines format to .codeweaver/progress.jsonl)
• Checkpoint/resume capability
• TypeScript strict mode, ESM modules
• Vitest test framework (73 tests passing)

🏗️ Architecture

Multi-Agent System

graph TB
    CLI[CLI Interface<br/>7 Command Groups] --> SERVICE[CodeWeaverService]
    MCP[MCP Server<br/>19 Tools] --> SERVICE

    SERVICE --> DISC[Discovery Agent]
    SERVICE --> CACHE[Cache Agent]
    SERVICE --> SNIP[Snippets Agent]
    SERVICE --> SYM[Symbols Agent]
    SERVICE --> SEARCH[Search Agent]
    SERVICE --> ANALYSIS[Analysis Agent]
    SERVICE --> SEMANTIC[Semantic Index Agent]
    SERVICE --> WATCHER[File Watcher Agent]
    SERVICE --> VCS[VCS Agent]
    SERVICE --> STORE[Symbol Storage]

    DISC --> GRADLE[Gradle Parser]
    CACHE --> SHA[SHA-256 Hashing]
    SNIP --> TOKEN[Token Counter]
    SYM --> PLUGINREG[Plugin Registry]
    PLUGINREG --> JAVAPLUGIN[Java Plugin<br/>java-parser]
    PLUGINREG --> TSPLUGIN[TypeScript Plugin<br/>typescript-estree]
    PLUGINREG --> JSPLUGIN[JavaScript Plugin<br/>typescript-estree]
    PLUGINREG --> MDPLUGIN[Markdown Plugin<br/>remark]
    PLUGINREG --> PYPLUGIN[Python Plugin<br/>tree-sitter WASM]
    SEARCH --> REGEX[Regex Matching]
    ANALYSIS --> COMPLEXITY[Cyclomatic<br/>Complexity]
    SEMANTIC --> LANCEDB[LanceDB<br/>Vector Search]
    SEMANTIC --> ONNX[ONNX Runtime<br/>Embeddings]
    WATCHER --> CHOKIDAR[Chokidar<br/>File Watching]
    WATCHER --> SEMANTIC
    VCS --> GIT[Git Operations]
    STORE --> JSONL[JSON Lines]

    style SERVICE fill:#e1f5ff
    style CLI fill:#d4edda
    style MCP fill:#fff3cd
    style SYM fill:#cfe2ff
    style SEARCH fill:#cfe2ff
    style ANALYSIS fill:#cfe2ff
    style SEMANTIC fill:#fff3cd
    style WATCHER fill:#fff3cd
    style VCS fill:#cfe2ff

Directory Structure

src/
├── index.ts                      # Main entry (auto-detection)
├── cli/
│   ├── index.ts                  # CLI entry point
│   └── commands/
│       ├── info.ts               # Info command
│       ├── file.ts               # File commands
│       ├── symbols.ts            # Symbols commands (Phase 2)
│       ├── search.ts             # Search commands (Phase 2)
│       └── analysis.ts           # Analysis commands (Phase 3)
├── mcp/
│   ├── index.ts                  # MCP entry point
│   ├── server.ts                 # MCPServer class
│   └── tools.ts                  # Tool registration (12 tools)
├── core/
│   ├── service.ts                # Shared business logic
│   ├── agents/
│   │   ├── discovery.ts          # Gradle analysis
│   │   ├── cache.ts              # Caching
│   │   ├── snippets.ts           # File reading
│   │   ├── symbols.ts            # Multi-language symbol extraction (Phase 2)
│   │   ├── search.ts             # Keyword/pattern search (Phase 2)
│   │   ├── analysis.ts           # Complexity analysis (Phase 3)
│   │   ├── semantic.ts           # Vector search (Phase 4)
│   │   ├── vcs.ts                # Git operations (Phase 4)
│   │   └── watcher.ts            # File watching (Phase 4)
│   ├── language/
│   │   ├── plugin.ts             # LanguagePlugin interface
│   │   ├── detector.ts           # Language detection
│   │   ├── registry.ts           # Plugin registry
│   │   └── plugins/
│   │       ├── java/
│   │       │   ├── index.ts      # JavaLanguagePlugin
│   │       │   ├── parser.ts     # java-parser wrapper
│   │       │   └── extractor.ts  # Java symbol extraction
│   │       └── typescript/
│   │           ├── index.ts      # TypeScript/JavaScriptLanguagePlugin
│   │           ├── parser.ts     # typescript-estree wrapper
│   │           └── extractor.ts  # TS/JS symbol extraction
│   └── storage/
│       └── json-symbol-store.ts  # Symbol index
├── types/
│   ├── mcp.ts
│   ├── progress.ts
│   ├── project.ts
│   ├── cache.ts
│   ├── symbols.ts
│   └── analysis.ts               # Analysis types (Phase 3)
└── utils/
    ├── progress-writer.ts        # Progress tracking
    └── mode-detector.ts          # CLI vs MCP detection

tests/
├── unit/                         # 184 passing tests
│   ├── mcp/server.test.ts        # 6 tests
│   ├── agents/
│   │   ├── discovery.test.ts     # 4 tests
│   │   ├── cache.test.ts         # 5 tests
│   │   ├── snippets.test.ts      # 7 tests
│   │   ├── symbols.test.ts       # 23 tests (Java)
│   │   ├── search.test.ts        # 11 tests
│   │   ├── analysis.test.ts      # 11 tests
│   │   ├── semantic.test.ts      # Tests for semantic search
│   │   ├── vcs.test.ts           # Tests for VCS operations
│   │   └── watcher.test.ts       # Tests for file watcher
│   ├── language/
│   │   ├── detector.test.ts      # Language detection tests
│   │   ├── registry.test.ts      # Plugin registry tests
│   │   ├── java.test.ts          # Java plugin tests
│   │   └── typescript.test.ts    # 21 tests (TypeScript/JavaScript)
│   └── storage/
│       └── json-symbol-store.test.ts  # 5 tests
├── integration/                  # 12 passing tests
│   ├── smoke.test.ts             # 5 tests
│   └── multi-language.test.ts    # 12 tests (Multi-Language Integration)
└── fixtures/
    ├── gradle-projects/simple/   # Gradle test fixtures
    ├── java/                     # Java test files
    └── typescript/               # TypeScript/JavaScript test files

📖 Usage

As CLI Tool

Project & Files:

# Show project information
codeweaver info

# Read entire file
codeweaver file read src/core/service.ts

# Read file with line numbers
codeweaver file read src/core/service.ts --numbers

# Read file with token limit
codeweaver file read src/core/service.ts --limit 500

# Read specific lines (1-indexed, inclusive)
codeweaver file range src/core/service.ts 10 20

# Get context around line (default: ±5 lines)
codeweaver file context src/core/service.ts 42
codeweaver file context src/core/service.ts 42 --context 10

Symbols (Phase 2):

# Index entire project
codeweaver symbols index

# Find symbols by name (case-insensitive)
codeweaver symbols find "UserService"
codeweaver symbols find "get"  # Finds all getXxx methods

# Get specific symbol by qualified name
codeweaver symbols get "com.example.UserService"
codeweaver symbols get "com.example.UserService#findById"

# List all symbols of a kind
codeweaver symbols list class
codeweaver symbols list method
codeweaver symbols list field

Search (Phase 2):

# Search for keyword
codeweaver search keyword "TODO"
codeweaver search keyword "processData"

# Case-insensitive search
codeweaver search keyword "exception" --case-insensitive
codeweaver search keyword "exception" -i

# Search with context lines
codeweaver search keyword "TODO" --context 3 -c 3

# Limit results
codeweaver search keyword "public" --max-results 10 -m 10

# Filter by file extensions
codeweaver search keyword "interface" --extensions .java .ts -e .java -e .ts

# Find files by pattern
codeweaver search files "*.java"
codeweaver search files "*Test.java"
codeweaver search files "User*.ts"

As MCP Server

1. Configure MCP Client

Add to your MCP configuration (e.g., Claude Desktop):

{
  "mcpServers": {
    "codeweaver": {
      "command": "node",
      "args": [
        "/absolute/path/to/mcp-workbench/dist/index.js",
        "--mcp"
      ],
      "cwd": "/path/to/your/java/project"
    }
  }
}

Or use npm:

{
  "mcpServers": {
    "codeweaver": {
      "command": "npm",
      "args": ["run", "dev", "--", "--mcp"],
      "cwd": "/absolute/path/to/mcp-workbench"
    }
  }
}

2. Available MCP Tools (10 total)

Project & Files:

project.meta - Get project metadata

// Input: {} (no parameters)
// Output: ProjectMetadata
{
  "name": "my-project",
  "version": "1.0.0",
  "javaVersion": "21",
  "gradleVersion": "8.5",
  "modules": [...],
  "dependencies": [...],
  "plugins": [...]
}

file.read - Read file with token limit

// Input: { filePath: string, maxTokens?: number }
await mcp.call('file.read', {
  filePath: 'src/main/java/com/example/App.java',
  maxTokens: 5000
});

file.readRange - Read specific lines

// Input: { filePath: string, startLine: number, endLine: number }
await mcp.call('file.readRange', {
  filePath: 'src/main/java/com/example/App.java',
  startLine: 10,
  endLine: 30
});

file.readWithNumbers - Read with line numbers

// Input: { filePath: string }
await mcp.call('file.readWithNumbers', {
  filePath: 'src/main/java/com/example/App.java'
});
// Output: "  1: package com.example;\n  2: \n  3: public class App { ... }"

Symbols (Phase 2):

symbols.index - Index entire project

// Input: {} (no parameters)
await mcp.call('symbols.index', {});
// Output: { files: 15, symbols: 234, classes: 12, classList: [...] }

symbols.find - Find symbols by name

// Input: { name: string }
await mcp.call('symbols.find', {
  name: 'UserService'
});
// Output: SymbolDefinition[]

symbols.findByKind - Find symbols by kind

// Input: { kind: 'class' | 'method' | 'field' | 'constructor' }
await mcp.call('symbols.findByKind', {
  kind: 'method'
});
// Output: SymbolDefinition[]

symbols.get - Get symbol by qualified name

// Input: { qualifiedName: string }
await mcp.call('symbols.get', {
  qualifiedName: 'com.example.UserService#findById'
});
// Output: SymbolDefinition

Search (Phase 2):

search.keyword - Search for keyword in files

// Input: { keyword: string, caseSensitive?: boolean, maxResults?: number, contextLines?: number, fileExtensions?: string[] }
await mcp.call('search.keyword', {
  keyword: 'TODO',
  caseSensitive: false,
  maxResults: 50,
  contextLines: 2,
  fileExtensions: ['.java', '.ts']
});
// Output: SearchResult[] with file, line, column, content, beforeContext, afterContext

search.files - Find files by pattern

// Input: { pattern: string }
await mcp.call('search.files', {
  pattern: '*Test.java'
});
// Output: string[] (file paths)

🧪 Testing

# Run all tests
npm test

# Run tests in CI mode (no watch)
npm test -- --run

# Run specific test file
npm test -- tests/unit/agents/snippets.test.ts

Test Coverage:

• ✅ MCP Server (6 tests)
• ✅ Discovery Agent (4 tests)
• ✅ Cache Agent (5 tests)
• ✅ Snippets Agent (7 tests)
• ✅ Symbol Storage (5 tests)
• ✅ Symbols Agent (23 tests - Java)
• ✅ Language Plugins (21 tests - TypeScript/JavaScript, 13 tests - Markdown)
• ⚠️ Python Plugin (18 tests - skipped due to WASM config)
• ✅ Search Agent (11 tests)
• ✅ Analysis Agent (11 tests)
• ✅ Semantic Agent (tests for vector search)
• ✅ VCS Agent (tests for Git operations)
• ✅ File Watcher Agent (tests for file watching)
• ✅ Integration Tests (17 tests: 5 smoke + 12 multi-language)
• Total: 256 passing (100%)

🔧 Development

Prerequisites

• Node.js >= 20.0.0
• TypeScript 5.7+
• Java JDK 21 (for target projects)
• Gradle (optional, wrapper preferred)

Setup

# Clone repository
git clone <repository-url>
cd mcp-workbench

# Install dependencies
npm install

# Build
npm run build

# Development mode (with auto-reload)
npm run build:watch

# Run in dev mode (no build required)
npm run dev

Scripts

npm run build           # Compile TypeScript
npm run build:watch     # Watch mode
npm run dev             # Run with tsx (no build)
npm test                # Run tests (watch mode)
npm run lint            # ESLint
npm run format          # Prettier
npm run clean           # Remove dist & cache
npm run validate-links  # Validate markdown links 🆕

Documentation Tools:

npm run validate-links           # Validate internal links (~2s)
npm run validate-links:external  # Include external links (~30s)
npm run validate-links:verbose   # Detailed output

See for details.

📊 Token Efficiency

CodeWeaver is designed to minimize token usage when providing code context to LLMs:

Strategies

1. Line Ranges: Only send requested line ranges, not entire files
2. Token Limits: Automatic truncation to configurable limits (default: 10k)
3. Smart Truncation: Respects word boundaries when truncating
4. Token Counting: Simple heuristic (~4 chars/token) for quick estimates
5. Context Windows: Provide minimal context around specific lines

Token Estimation

Max Response Size: 10,000 tokens (~40 KB text)

🗺️ Roadmap

✅ Phase 1: Foundation (Complete - 100%)

• ✅ MCP Server skeleton with tool registration
• ✅ Progress tracking (JSON Lines)
• ✅ Discovery Agent (Gradle metadata)
• ✅ Cache Agent (content-addressable storage)
• ✅ Symbol Storage (JSON Lines persistence)
• ✅ Core Service (shared logic)
• ✅ CLI Interface with commands
• ✅ MCP Interface with stdio
• ✅ Build & Test Setup (32 tests passing)
• ✅ Snippets Agent with token limits
• ✅ Documentation (complete)
• ✅ Integration Tests (5 smoke tests)

✅ Phase 2: Indexing (Complete - 100%)

• ✅ Symbols Agent (java-parser, symbol extraction)
• ✅ Search Agent (keyword + pattern search)
• ✅ Project-wide indexing (classes, methods, fields, constructors)
• ✅ Symbol search (by name, kind, qualified name)
• ✅ File search (glob patterns with * and ?)
• ✅ Context search (lines before/after matches)
• ✅ MCP Tools integration (6 new tools)
• ✅ CLI Commands integration (symbols, search)
• ✅ Full test coverage (19 new tests)
• ✅ Documentation update

Note: LanceDB semantic search deferred to later phase as enhancement

✅ Phase 3: Analysis (Complete - 100%)

• ✅ Analysis Agent (complexity & metrics calculation)
• ✅ Cyclomatic Complexity calculation (if, loops, catch, &&, ||, ?:)
• ✅ Code Metrics (LOC, SLOC, comments, blank lines)
• ✅ Import analysis
• ✅ Method call detection
• ✅ Project-wide statistics (total complexity, average, top N files)
• ✅ MCP Tools integration (2 new tools)
• ✅ CLI Commands integration (analysis)
• ✅ Full test coverage (11 new tests)
• ✅ Documentation update

Note: Static analysis tools (SpotBugs, Checkstyle) and Gradle runner deferred

✅ Phase 4: VCS Integration (Complete - 100%)

• ✅ VCS Agent (Git operations)
• ✅ Repository status (modified, added, deleted, untracked files)
• ✅ Diff generation (file-level and project-level)
• ✅ Blame information (line-by-line authorship)
• ✅ Commit history (with filtering options)
• ✅ Branch management (list, compare)
• ✅ MCP Tools integration (6 new tools)
• ✅ CLI Commands integration (vcs)
• ✅ Full test coverage (11 new tests)
• ✅ Documentation update

📋 Phase 5: Orchestration (Planned)

• Orchestrator Agent (DAG-based pipeline)
• Parallel task execution
• Dependency resolution

🐛 Troubleshooting

Tests failing?

# Clean and reinstall
npm run clean
rm -rf node_modules package-lock.json
npm install
npm test -- --run

Build errors?

# Check TypeScript version
npx tsc --version  # Should be 5.7+

# Rebuild
npm run clean
npm run build

MCP server not responding?

# Check if running in MCP mode
npm run dev -- --mcp

# Verify stdio transport
echo '{}' | npm run dev -- --mcp

CLI not working?

# Ensure TTY mode (not piped)
npm run dev -- info

# Check built binary
node dist/index.js info

📝 Progress Tracking

View live progress during implementation:

# Bash/Git Bash
tail -f .codeweaver/progress.jsonl

# PowerShell
Get-Content .codeweaver\progress.jsonl -Wait

# Read checkpoint
cat .codeweaver/checkpoint.json

🤝 Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Write tests for new features
4. Ensure all tests pass: npm test -- --run
5. Follow code style (ESLint + Prettier)
6. Submit a pull request

Code Style

• TypeScript Strict Mode: All type errors must be resolved
• ESM Modules: Use .js extensions in imports
• Test-Driven Development: Write tests first
• No Unused Variables: Clean code, no warnings

📄 License

MIT License - see

📚 Documentation

Vollständige Dokumentation in

🗺️ Schnellzugriff

Komplette Navigation: - Vollständiger Dokumentations-Index mit Navigation nach Rolle

Beliebte Dokumente:

• - 🚀 START HERE! Complete Developer Workflow (All 19 MCP tools in action)
• - ⚡ 5-Minuten Quick-Start Guide
• - 📚 Complete API Reference (alle 19 MCP Tools)
• - ⚡ Performance Benchmarks & Scalability
• - 🔥 Semantic Search Guide mit Workflows
• - 🏗️ System-Architektur (Dual-Mode, Multi-Agent)
• - 📖 Glossar aller Begriffe & Akronyme

🔗 Externe Links

• Model Context Protocol - MCP Specification

🎯 Current Status

Alpha Release v0.1.0 ⚠️

✅ Working Features:

• MCP Server with 19 tools (project, files, symbols, search, analysis, vcs)
• CLI with 7 command groups (includes watch mode)
• Multi-Language Support - Java, TypeScript, JavaScript, Markdown, Python with plugin architecture
• Semantic Search with ONNX Runtime optimizations
• Multi-Collection Support (Code + Docs)
• File Watcher for automatic index updates
• Symbol Extraction - Complete support for Java, TypeScript, JavaScript, Markdown, and Python
• Code Quality Analysis - Cyclomatic complexity, LOC metrics
• Git Integration - Status, diff, blame, log, branches
• 291 tests passing (100% - all features tested)

⚠️ Known Limitations:

• Performance varies on large codebases (>10k files)
• Semantic search memory usage can be high
• File watcher may miss rapid changes
• Documentation is incomplete
• Breaking changes expected in future releases

🔮 Planned Improvements:

• Python WASM configuration ✅ COMPLETED
• GPU acceleration for semantic search
• Better error messages
• More language support (Go, Rust, C#, etc.)
• Performance optimizations
• Comprehensive documentation

💡 Philosophy

CodeWeaver follows these principles:

1. Token Efficiency First - Never overwhelm LLMs with entire files
2. Zero Native Dependencies - Pure Node.js for portability
3. Test-Driven Development - Tests before implementation
4. Dual Interface - Same codebase serves CLI and MCP
5. Progressive Enhancement - Working foundation, build up from there

Built with ❤️ for the LLM-assisted development workflow.