代理控制框架
HTTP-SSESTDIO全面自主代理框架，具备64+专业工具
全面自主代理框架，具备64+专业工具
A comprehensive toolkit for autonomous agent development with 64+ specialized tools
The Agentic Control Framework (ACF) is a production-ready platform that transforms your existing projects into powerful autonomous agents. With 64+ specialized tools spanning task management, filesystem operations, browser automation, terminal control, and more, ACF provides everything needed to build sophisticated AI agents.
Key Features:
agentic-control-framework/
├── 📁 bin/ # CLI executables and entry points
│ ├── 🔧 task-manager # Main CLI entry point for task management
│ ├── 🔧 task-manager-mcp # Legacy MCP server wrapper (deprecated)
│ ├── 🔧 agentic-control-framework-mcp # Primary MCP server for Cursor/Claude integration
│ └── 🔧 task-cli.js # Alternative CLI interface with enhanced features
│
├── 📁 src/ # Core source code
│ ├── 🔧 core.js # Core task management logic and data structures
│ ├── 🔧 cli.js # CLI command definitions and argument parsing
│ ├── 🔧 mcp_server.js # MCP server with 64+ tools (main server file)
│ ├── 🔧 filesystem_tools.js # Filesystem operations for MCP integration
│ ├── 🔧 prd_parser.js # AI-powered PRD parsing and task generation
│ ├── 🔧 tableRenderer.js # Task table rendering and formatting utilities
│ ├── 🔧 logger.js # Standardized logging across all components
│ │
│ └── 📁 tools/ # Additional test utilities and helpers
│ ├── 🔧 browser_tools.js # Browser automation with Playwright
│ ├── 🔧 terminal_tools.js # Command execution, process management
│ ├── 🔧 search_tools.js # Advanced code search with ripgrep
│ ├── 🔧 edit_tools.js # Surgical text editing and replacement
│ ├── 🔧 applescript_tools.js # macOS automation and system integration
│ └── 🔧 enhanced_filesystem_tools.js # Extended filesystem operations
│
├── 📁 test/ # Testing infrastructure
│ ├── 🔧 comprehensive_mcp_test.js # Consolidated test suite for all MCP tools
│ ├── 🔧 test_env_guardrails.js # Security and environment validation tests
│ ├── 🔧 test_filesystem.js # Filesystem operations testing
│ └── 📁 scripts/ # Additional test utilities and helpers
│
├── 📁 docs/ # Documentation and guides
│ ├── 📘 tutorial.md # Step-by-step getting started guide
│ ├── 📘 MCP_INTEGRATION_GUIDE.md # Detailed MCP setup and configuration
│ ├── 📘 enhanced-mcp-tools.md # Complete tool reference documentation
│ ├── 📘 COMPLETE_TUTORIAL.md # Comprehensive usage documentation
│ ├── 📁 tutorials/ # Step-by-step implementation guides
│ └── 📁 mcp-integration/ # MCP protocol documentation and examples
│
├── 📁 tasks/ # Generated task files and project data
│ ├── 📄 Individual task Markdown files (auto-generated)
│ └── 📄 Task status and progress tracking files
│
├── 📁 templates/ # Project templates and scaffolding
│ ├── 📄 PRD templates for different project types
│ ├── 📄 Project initialization templates
│ └── 📄 Task structure templates
│
├── 📁 scripts/ # Automation and utility scripts
│ ├── 🔧 sync-upstream.sh # Sync with upstream MCP repositories
│ ├── 🔧 compare-upstream-tools.js # Compare tool versions with upstream
│ └── 📄 Various maintenance and setup scripts
│
├── 📁 .github/ # GitHub Actions and repository configuration
│ ├── 📁 workflows/ # CI/CD pipelines and automation
│ └── 📄 Issue and PR templates
│
├── 📁 client-configurations/ # MCP client setup examples
│ ├── 📄 Cursor configuration examples
│ ├── 📄 Claude Desktop setup files
│ └── 📄 Various IDE integration examples
│
├── 📁 deployment/ # Cloud deployment configurations
│ ├── 📄 GCP deployment configurations
│ ├── 📄 Railway deployment files
│ └── 📄 Docker configurations
│
├── 📁 public/ # Static assets and public files
│ └── 📄 Web interface assets (for cloud deployment)
│
├── 📁 .tasks/ # Internal task management data
│ └── 📄 Task storage and metadata (auto-managed)
│
├── 📁 .tmp/ # Temporary files and cache
│ └── 📄 Temporary workspace data and cache files
│
├── 📁 MCP documents/ # MCP protocol documentation and examples
│ ├── 📄 Protocol specifications
│ ├── 📄 Integration examples and best practices
│ └── 📄 Client setup instructions
│
├── 📄 Core Configuration Files
├── 📄 package.json # Node.js dependencies and project metadata
├── 📄 config.json # ACF configuration and settings
├── 📄 tasks.json # Main task data storage
├── 📄 mcp-connection.json # MCP connection configuration
├── 📄 settings.json # User preferences and customizations
├── 📄 env.example # Environment variables template
├── 📄 setup.sh # Automated setup and installation script
├── 📄 quick-deploy.sh # One-command deployment script
│
├── 📄 Testing & Validation Files
├── 📄 test-simple-tools.js # Lightweight test suite for all tools
├── 📄 test-all-tools-comprehensive.sh # Comprehensive bash-based test suite
├── 📄 test-mcp-proxy.sh # mcp-proxy integration testing
├── 📄 test-mcp-clients.sh # Client configuration testing
├── 📄 test-deployment-complete.sh # End-to-end deployment validation
│
├── 📄 Deployment Files
├── 📄 Dockerfile.proxy # Docker configuration for mcp-proxy
├── 📄 docker-compose.yml # Multi-container deployment
├── 📄 mcp-proxy-config.yaml # mcp-proxy configuration
├── 📄 auth-proxy.js # Authentication proxy for cloud deployment
│
├── 📄 Documentation Files
├── 📄 README.md # This comprehensive guide
├── 📄 CHANGES.md # Version history and changelog
├── 📄 ACF-TESTING-SUMMARY.md # Latest test analysis and status
├── 📄 SIMPLE-TEST-REPORT.md # Automated test execution results
├── 📄 COMPREHENSIVE-TEST-REPORT.md # Detailed testing documentation
├── 📄 DEPLOYMENT-STATUS.md # Current deployment status and guides
├── 📄 GCP-DEPLOYMENT-GUIDE.md # Google Cloud Platform deployment
├── 📄 MCP-PROXY-DEPLOYMENT.md # mcp-proxy setup and configuration
├── 📄 CURSOR-SETUP-GUIDE.md # Cursor IDE integration guide
├── 📄 SETUP-INSTRUCTIONS.md # Quick setup instructions
├── 📄 WORKING-EXAMPLE.md # Live examples and demonstrations
├── 📄 MANUS_LIKE_ENHANCEMENT_PLAN.md # Roadmap for autonomous agent features
├── 📄 CONSOLIDATED_TEST_REPORTS.md # Consolidated testing analysis
│
└── 📄 tasks-table.md # Human-readable task status table
task-manager: Main CLI entry point for task managementagentic-control-framework-mcp: Primary MCP server for Cursor/Claude integrationtask-cli.js: Alternative CLI interface with enhanced featurescore.js: Core task management logic and data structuresmcp_server.js: MCP server with 64+ tools (main server file)filesystem_tools.js: Filesystem operations for MCP integration| Component | Status | Details |
|---|---|---|
| CLI Mode | ✅ 100% Working | All task management and core tools functional |
| Local MCP | ✅ 100% Working | All tools functional - terminal, search, filesystem fixed |
| Cloud MCP | ✅ Ready | mcp-proxy integration working, startup timing optimized |
| Browser Tools | ✅ 100% Ready | Playwright integration complete |
| AppleScript | ✅ 100% Ready | macOS automation ready |
All tests passing! See ACF-TESTING-SUMMARY.md for detailed test results
# Install Node.js 18+ node --version # Install dependencies npm install # Make CLI tools executable chmod +x bin/*
# Run comprehensive test suite node test-simple-tools.js
Perfect for: Automated scripts, local development, CI/CD integration
# Initialize project cd your-project ./path/to/acf/bin/task-manager init -n "My Project" -d "Project description" # Add tasks ./path/to/acf/bin/task-manager add -t "Implement feature" -d "Add new functionality" -p high # List tasks ./path/to/acf/bin/task-manager list # Update task status ./path/to/acf/bin/task-manager status 1 inprogress -m "Started working" # Add subtasks ./path/to/acf/bin/task-manager add-subtask 1 -t "Write tests" # Get next actionable task ./path/to/acf/bin/task-manager next # Generate task files ./path/to/acf/bin/task-manager generate
# Update task details ./path/to/acf/bin/task-manager update 1 -p medium --related-files "src/main.js,test/main.test.js" # Get task context ./path/to/acf/bin/task-manager get-context 1 # Remove completed tasks ./path/to/acf/bin/task-manager remove 1 # Generate markdown table ./path/to/acf/bin/task-manager table
# Daily standup automation #!/bin/bash echo "📊 Daily Standup Report" echo "=======================" ./bin/task-manager list --status inprogress echo "" echo "Next Priority Tasks:" ./bin/task-manager next # CI/CD Integration #!/bin/bash # In your CI pipeline ./bin/task-manager add -t "Deploy v$VERSION" -d "Deploy to production" -p high ./bin/task-manager status $TASK_ID done -m "Deployed successfully"
Perfect for: IDE integration (Cursor, Claude Desktop), local development
acf-localnode["/path/to/agentic-control-framework/bin/agentic-control-framework-mcp", "--workspaceRoot", "/path/to/your/project"]{ "WORKSPACE_ROOT": "/path/to/your/project", "ALLOWED_DIRS": "/path/to/your/project:/tmp", "READONLY_MODE": "false" }
{ "mcp.servers": { "acf-local": { "command": "node", "args": [ "/path/to/agentic-control-framework/bin/agentic-control-framework-mcp", "--workspaceRoot", "/path/to/your/project" ], "env": { "WORKSPACE_ROOT": "/path/to/your/project", "ALLOWED_DIRS": "/path/to/your/project:/tmp", "READONLY_MODE": "false" } } } }
{ "mcpServers": { "acf-local": { "command": "node", "args": [ "/path/to/agentic-control-framework/bin/agentic-control-framework-mcp", "--workspaceRoot", "/path/to/your/project" ], "env": { "WORKSPACE_ROOT": "/path/to/your/project", "ALLOWED_DIRS": "/path/to/your/project:/tmp" } } } }
Once configured, you can use natural language with your AI assistant:
"Add a new high-priority task for implementing user authentication"
"List all tasks that are currently in progress"
"Read the contents of src/main.js and create a task for adding error handling"
"Execute the test suite and create a task if there are failures"
"Search for all TODO comments in the codebase and create tasks for them"
"Take a screenshot of the application login page"
"Write a new file called docs/api.md with API documentation"
| Category | Tools | Status |
|---|---|---|
| Task Management | listTasks, addTask, updateStatus, getNextTask, etc. | ✅ Working |
| Filesystem | read_file, write_file, list_directory, search_files | ✅ Working |
| Terminal | execute_command, list_processes, kill_process | ⚠️ Partial |
| Browser | navigate, click, type, screenshot, pdf_save | ✅ Working |
| Search/Edit | search_code, edit_block | ⚠️ Partial |
| AppleScript | applescript_execute (macOS only) | ✅ Working |
Perfect for: Remote access, web clients, multi-client support
# Install mcp-proxy npm install -g mcp-proxy # Start ACF with mcp-proxy export WORKSPACE_ROOT="/path/to/your/project" export ALLOWED_DIRS="/path/to/your/project:/tmp" mcp-proxy --port 8080 \ node bin/agentic-control-framework-mcp \ --workspaceRoot "$WORKSPACE_ROOT"
# Health check curl http://localhost:8080/health # MCP initialization curl -X POST http://localhost:8080/stream \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0.0"}}}' # List available tools curl -X POST http://localhost:8080/stream \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}' # Call a tool curl -X POST http://localhost:8080/stream \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"listTasks","arguments":{}}}'
{ "mcp.servers": { "acf-cloud": { "transport": "sse", "endpoint": "http://localhost:8080/sse" } } }
# Authenticate gcloud auth login # Create project gcloud projects create acf-your-name-$(date +%s) export GCP_PROJECT_ID="your-project-id" # Deploy ./quick-deploy.sh gcp --proxy-only
Task Management:
- listTasks: List all tasks with filtering
- addTask: Create new tasks with priority/dependencies
- addSubtask: Add subtasks to existing tasks
- updateStatus: Change task status (todo/inprogress/done/blocked/error)
- updateTask: Modify task details
- removeTask: Delete tasks or subtasks
- getNextTask: Get next actionable task based on priority/dependencies
- getContext: Get detailed task information
- generateTaskFiles: Create individual markdown files for tasks
- generateTaskTable: Create readable task status table
- parsePrd: Parse Product Requirements Documents
- expandTask: AI-powered task breakdown
- reviseTasks: AI-powered task revision
- setWorkspace: Configure workspace directory
- initProject: Initialize new ACF project
File Operations:
- read_file: Read file contents with type detection
- read_multiple_files: Read multiple files at once
- write_file: Write/create files
- copy_file: Copy files and directories
- move_file: Move/rename files
- delete_file: Delete files/directories
- get_file_info: File metadata and statistics
Directory Operations:
- list_directory: Detailed directory listing
- create_directory: Create directories
- tree: Hierarchical directory structure
- search_files: Find files by pattern
- list_allowed_directories: Show accessible paths
- get_filesystem_status: Show security status
Command Execution:
- execute_command: Run shell commands with timeout
- read_output: Read from running processes
- force_terminate: Kill processes
- list_sessions: Show active terminal sessions
- list_processes: Show running processes
- kill_process: Terminate processes
Navigation:
- browser_navigate: Navigate to URLs
- browser_navigate_back: Go back
- browser_navigate_forward: Go forward
- browser_close: Close browser
Interaction:
- browser_click: Click elements
- browser_type: Type text
- browser_hover: Hover over elements
- browser_drag: Drag and drop
- browser_select_option: Select dropdown options
- browser_press_key: Keyboard input
Capture:
- browser_take_screenshot: Screenshots
- browser_snapshot: Accessibility snapshots
- browser_pdf_save: Save as PDF
Management:
- browser_tab_list: List browser tabs
- browser_tab_new: Open new tabs
- browser_tab_select: Switch tabs
- browser_tab_close: Close tabs
- browser_file_upload: Upload files
- browser_wait: Wait for time/conditions
- browser_resize: Resize window
- browser_handle_dialog: Handle alerts/dialogs
- browser_console_messages: Get console logs
- browser_network_requests: Monitor network
Code Operations:
- search_code: Advanced text/code search with ripgrep
- edit_block: Surgical text replacements
macOS Automation:
- applescript_execute: Run AppleScript for system integration
Server Management:
- get_config: Get server configuration
- set_config_value: Update configuration values
# CLI approach ./bin/task-manager init -n "E-commerce App" -d "Build online store" ./bin/task-manager add -t "Setup project structure" -p high ./bin/task-manager add -t "Configure database" -p high ./bin/task-manager add -t "Implement user auth" -p medium ./bin/task-manager add -t "Add payment integration" -p medium ./bin/task-manager add -t "Deploy to production" -p low
// MCP approach - ask your AI assistant: "Search the codebase for any TODO comments and create tasks for each one" "Read all JavaScript files in src/ and create tasks for any functions missing error handling" "Take a screenshot of the app and create a task for any UI issues you notice"
#!/bin/bash # In your GitHub Actions workflow - name: Update project tasks run: | ./bin/task-manager add -t "Test release v${{ github.event.release.tag_name }}" -p high ./bin/task-manager status $TASK_ID inprogress -m "Running tests for ${{ github.sha }}" # Run tests npm test if [ $? -eq 0 ]; then ./bin/task-manager status $TASK_ID done -m "Tests passed" else ./bin/task-manager status $TASK_ID error -m "Tests failed" fi
// Via MCP in your IDE "Navigate to our staging site and take screenshots of the login, dashboard, and profile pages" "Fill out the contact form with test data and take a screenshot of the success page" "Test the mobile responsiveness by resizing to phone dimensions and taking screenshots"
# Comprehensive test suite node test-simple-tools.js # Individual component tests ./test-all-tools-comprehensive.sh
# Clone repository git clone https://github.com/your-org/agentic-control-framework.git cd agentic-control-framework # Install dependencies npm install # Setup development environment chmod +x bin/* export WORKSPACE_ROOT="$(pwd)" export ALLOWED_DIRS="$(pwd):/tmp" # Test CLI mode ./bin/task-manager list # Test MCP mode node bin/agentic-control-framework-mcp
# Check if tasks.json exists ls -la tasks.json # Verify permissions chmod +x bin/task-manager # Check Node.js version node --version # Should be 18+
# Check environment variables echo $WORKSPACE_ROOT echo $ALLOWED_DIRS # Test MCP server directly node bin/agentic-control-framework-mcp --help # Check file permissions ls -la bin/agentic-control-framework-mcp
# Check mcp-proxy installation npm list -g mcp-proxy # Test proxy health curl http://localhost:8080/health # Check proxy logs mcp-proxy --port 8080 --debug node bin/agentic-control-framework-mcp
git checkout -b feature/amazing-featurenode test-simple-tools.jsgit commit -m 'Add amazing feature'git push origin feature/amazing-featureThis project is licensed under the MIT License - see the LICENSE file for details.
🚀 Ready to build your autonomous agent? Choose your mode and get started!
| Mode | Use Case | Setup Time | Status |
|---|---|---|---|
| CLI | Scripts, automation | 2 minutes | ✅ Production Ready |
| Local MCP | IDE integration | 5 minutes | ✅ Production Ready |
| Cloud MCP | Remote access | 15 minutes | ✅ Production Ready |
For detailed test results and improvement roadmap, see ACF-TESTING-SUMMARY.md.