
ZMCPTools
STDIOTypeScript MCP server with 61 tools for multi-agent orchestration and documentation intelligence
TypeScript MCP server with 61 tools for multi-agent orchestration and documentation intelligence
🚀 TypeScript MCP Tools for Claude Code - Professional multi-agent orchestration platform with 61 enhanced tools, documentation intelligence, and advanced automation capabilities.
Before spawning agents, run this command once to enable proper agent permissions:
claude --dangerously-skip-permissions
Agents run on daemon threads and need this permission to execute properly.
claude-mcp-tools
and claude-mcp-server
commands# Install globally first (recommended for WSL/Linux compatibility) pnpm add -g zmcp-tools # If requested, approve build scripts for native dependencies pnpm approve-builds -g # Then install MCP integration zmcp-tools install # Alternative: Direct installation (may have issues with Sharp in WSL) npx zmcp-tools@latest install # npm yarn dlx zmcp-tools@latest install # yarn bunx zmcp-tools@latest install # bun
This automatically:
~/.mcptools/server/
claude mcp add --scope local
(current directory only)# Clone and setup development environment git clone https://github.com/zachhandley/ZMCPTools cd ZMCPTools # Quick automated setup pnpm install && pnpm run install:global # Or manual setup pnpm install # Install dependencies pnpm build # Compile TypeScript pnpm link --global # Create global symlink zmcp-tools install # Configure MCP integration
Development features:
zmcp-tools
commandpnpm dev
pnpm build
pnpm test
The installer automatically configures the MCP server using claude mcp add --scope local
. The server runs directly with Node.js:
# Automatically executed during installation claude mcp add --scope local zmcp-tools node ~/.mcptools/server/index.js
This provides:
ZMCPTools TypeScript requires the following:
npm install -g typescript
)Note: This TypeScript implementation includes native LanceDB vector database with no Python dependencies required.
Here are some common workflows you can achieve with ZMCPTools:
"Using the zmcp server, find the styles and generate me a brand guide called THEWEBSITE_BRAND.md"
"Scrape https://modelcontextprotocol.io/introduction -- use the selector #content-area (by ID), don't allow any subdomains, and nothing ignored (though we can ask it to ignore regex, glob, patterns to *not* get some docs), update it weekly"
"Create a multi-agent orchestration to design and architect a ModelContextProtocol TypeScript server to do XYZ"
"Analyze the project structure, and then search the .treesummary directory to see what's there"
"Spawn an agent to do X, and then monitor its progress"
"Search the documentation for X"
"Create a browser session, navigate to https://example.com, take a screenshot, and analyze the page structure for accessibility issues"
"Store this implementation pattern in the knowledge graph and find related patterns we've used before"
"Create a multi-agent team with foundation session 'auth-refactor-2024' to refactor authentication across frontend and backend with 90% cost savings"
"Start a dev server, run the test suite, and spawn an agent to fix any failing tests while monitoring progress in real-time"
"Analyze recent agent errors, identify patterns, and spawn a debugging agent that learns from previous failures"
"Create an execution plan for implementing OAuth, then execute it with coordinated agents following the plan"
The Plan system provides 4 streamlined tools for orchestration:
create_execution_plan
- Create high-level execution plans from objectivesget_execution_plan
- Retrieve plans with progress tracking via linked Tasksexecute_with_plan
- Execute plans by creating coordinated Tasks for agentslist_execution_plans
- List and monitor execution plansPlans create Tasks for implementation - Plans are high-level orchestration templates while Tasks are specific work items assigned to agents.
ZMCPTools features an AI architect that automatically analyzes objectives and spawns coordinated agent teams with proper dependencies and real-time communication.
# Show help and available commands zmcp-tools --help # Show system status zmcp-tools status # Start the MCP server zmcp-server # Agent management zmcp-tools agent list zmcp-tools agent spawn -t <type> -r <repository> -d <description> zmcp-tools agent terminate -i <agent-id> # Task management zmcp-tools task list zmcp-tools task create -t <title> -d <description> # Memory operations zmcp-tools memory search -q <query> zmcp-tools memory store -t <title> -c <content> # Communication rooms zmcp-tools room list zmcp-tools room join -n <name>
# Initial setup (one time - using pnpm) pnpm install # Install dependencies first pnpm run install:global # Build, link globally, and configure everything # Alternative package managers npm install && npm run build && npm link && zmcp-tools install yarn install && yarn build && yarn link && zmcp-tools install bun install && bun run build && bun link && zmcp-tools install # Development with hot-reload pnpm dev # Start MCP server with tsx pnpm dev:cli # Start CLI with tsx # Building and testing pnpm build # Compile TypeScript to dist/ pnpm test # Run Vitest tests pnpm test:ui # Run tests with UI pnpm test:run # Run tests once # Code quality pnpm lint # ESLint checking pnpm typecheck # TypeScript type checking # Production pnpm start # Start compiled MCP server pnpm start:cli # Start compiled CLI # Management zmcp-tools install # Install/reinstall MCP server zmcp-tools uninstall # Remove MCP server and settings zmcp-tools status # Check system status zmcp-tools help # Show all commands # For users who installed via npx npx zmcp-tools@latest status # Check status npx zmcp-tools@latest uninstall # Remove installation
🌟 TypeScript Features:
🎯 Type Safety
🚀 Performance
🧪 Testing
📦 Module System
The following aliases are available (add to ~/.zshrc
):
alias mcp-tools="zmcp-tools" alias mcp-server="zmcp-server" alias mcp-status="zmcp-tools status" alias mcp-dev="npm run dev"
# Build configuration in tsconfig.json { "compilerOptions": { "target": "ES2022", "module": "ESNext", "moduleResolution": "bundler", "strict": true, "experimentalDecorators": true, "outDir": "dist", "rootDir": "src" } } # Development scripts (works with npm/yarn/pnpm/bun) npm run dev # Hot-reload development npm run build # Production build npm test # Run test suite
🎯 TypeScript Features:
The installer automatically configures the MCP server, but if you need to manually configure it:
# For production install (via npx) # Server is installed at ~/.mcptools/server/index.js # Configuration is automatic via 'claude mcp add --scope local' # For development install claude mcp add zmcp-tools $(pwd)/dist/server/index.js # Verify installation claude mcp list # Test server directly echo '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | node ~/.mcptools/server/index.js
Full TypeScript implementation with MCP SDK, proper error handling, and tool management.
Layer 1: Enhanced File Operations
Layer 2: Multi-Agent Orchestration
Automatic CLAUDE.md Integration:
<!-- zzZMCPToolszz START/END -->
Automatic Claude Hooks Integration:
Per-Project Setup:
# Automatic integration during project setup zmcp-tools install --project # Creates/updates: # • MCP server registration via 'claude mcp add --scope local' # • ./.claude/commands/ (Claude commands) # • ./.claude/hooks/ (session start context injection) # • ./.claude/settings.json (hook configuration) # • ./CLAUDE.md (integration guide with architect examples)
./dist/
)~/.mcptools/data/
(SQLite databases)~/.mcptools/lancedb/
(LanceDB vector database)~/.mcptools/data/claude_mcp_tools.db
# Clone and setup development environment git clone https://github.com/zachhandley/ZMCPTools cd ZMCPTools pnpm install # Quick setup pnpm run install:global && mcp-tools install # Build, link # Development mode pnpm dev # Run MCP server with hot-reload pnpm dev:cli # Run CLI with hot-reload # Build and test pnpm build # Compile TypeScript pnpm test # Run test suite # Test the binaries node dist/index.js # MCP server node dist/cli/index.js # CLI interface
$ zmcp-tools status
ZMCPTools Status:
✅ TypeScript Build: dist/ directory exists
✅ Data Directory: ~/.mcptools/data/
✅ SQLite Database: claude_mcp_tools.db
✅ LanceDB Vector Database: ~/.mcptools/lancedb/
✅ MCP Server: zmcp-server binary available
✅ Dependencies: @modelcontextprotocol/sdk, @lancedb/lancedb, better-sqlite3
$ npm run dev
Starting TypeScript development server...
✅ TypeScript compilation successful
✅ MCP server starting on stdio
✅ SQLite databases initialized
✅ LanceDB vector database initialized
✅ Agent orchestration ready
✅ Foundation cache system active
Listening for MCP requests...
Press Ctrl+C to stop
# Check prerequisites node --version # Node.js 18+ required pnpm --version # Package manager (or npm/yarn/bun) claude --version # Claude CLI required # Clean installation rm -rf node_modules dist pnpm install pnpm build # Development installation git clone https://github.com/zachhandley/ZMCPTools cd ZMCPTools pnpm install && pnpm run install:global
# Check build output ls -la dist/ node dist/index.js --help # Check data directory ls -la ~/.mcptools/data/ # Test MCP server claude mcp list echo '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | node dist/index.js
# Test TypeScript compilation pnpm typecheck pnpm lint # Test MCP server directly node dist/index.js # Debug with development server pnpm dev # Check MCP configuration claude mcp list claude mcp remove zmcp-tools claude mcp add zmcp-tools $(pwd)/dist/index.js
# Type checking errors pnpm typecheck # Check TypeScript errors npx tsc --noEmit --pretty # Detailed type errors # Runtime errors node --inspect dist/index.js # Debug with Node.js inspector pnpm dev # Hot-reload development # Database issues rm -rf ~/.mcptools/data/*.db # Reset databases node dist/index.js # Reinitialize # Dependency issues rm -rf node_modules pnpm-lock.yaml pnpm install # Clean dependency install
TypeScript-First Design:
Database Layer:
MCP Compliance:
git checkout -b feature/amazing-feature
)MIT License - see LICENSE file for details.
🚀 Supercharge your Claude Code workflows with ZMCPTools - TypeScript-powered multi-agent orchestration, LanceDB vector search, type-safe development, enhanced performance, and intelligent development assistance!