Portable MCP Toolkit

# MCP Tools Reference ## Available Tools ### Filesystem Server **Purpose**: Read, write, search files and directories **Common operations:** - `read_file(path)` - Read file contents - `write_file(path, content)` - Write/update file - `list_directory(path)` - List directory contents - `search_files(pattern)` - Find files matching pattern - `file_tree(path)` - Get directory structure **Usage in Cursor:** - "Read src/TaskOrchestrator.cs" - "Search for all files containing 'error handling'" - "Show directory structure of src/" ### Git Server **Purpose**: Version control operations **Common operations:** - `git_log(file)` - View commit history - `git_diff(commit1, commit2)` - Compare versions - `git_status()` - Check working directory status - `git_branch()` - List branches **Usage in Cursor:** - "Show git history for TaskOrchestrator.cs" - "What changed in the last 5 commits?" - "Show diff between main and feature branch" ### GitHub Server **Purpose**: GitHub API integration **Common operations:** - `list_prs(repo)` - List pull requests - `get_pr(number)` - Get PR details - `list_issues(repo)` - List issues - `search_code(query)` - Search GitHub code **Usage in Cursor:** - "Show open PRs for this repository" - "Get details of PR #42" - "Search for similar implementations on GitHub" ### Sequential Thinking **Purpose**: Complex multi-step reasoning **Usage in Cursor:** - "Use sequential-thinking to plan refactoring TaskOrchestrator into smaller services" - "Break down this complex feature into implementation steps" - "Analyze the architectural implications of this change" ### Brave Search (Optional) **Purpose**: Web search for research **Usage in Cursor:** - "Search for best practices in async error handling" - "Find recent articles about Result<T> pattern in C#" - "Research modern dependency injection patterns" ### Memory (Optional) **Purpose**: Context persistence across sessions **Usage in Cursor:** - "Remember that I prefer Result<T> over exceptions" - "Recall our discussion about async patterns" - "What preferences have you learned about my code style?" ## AIStack Intelligence Server (Production) **Purpose**: Local intelligence layer with hybrid architecture (FREE local processing + CHEAP remote generation) ### Semantic Code Search **Purpose**: Vector search over codebase using Qdrant (90% token reduction) **Common operations:** - `search_code_semantic(query, max_results, min_score)` - Semantic code search - `search_documentation(query, max_results)` - RAG over documentation **Usage in Cursor:** - "Use aistack-intelligence to search for error handling patterns" - "Find similar implementations of async cancellation" - "Search documentation for LangGraph checkpointing" **Token Savings**: 90% (500 vs 5000 tokens) ### Pattern Analysis **Purpose**: Analyze codebase patterns using local Ollama LLM (95% token reduction) **Common operations:** - `analyze_code_patterns(pattern_type, max_examples)` - Pattern analysis **Usage in Cursor:** - "Analyze error handling patterns in the codebase" - "What dependency injection patterns are used?" - "Show me async/await patterns" **Token Savings**: 95% (200 vs 4000 tokens) ### Impact Analysis **Purpose**: Multi-layer change impact analysis (85% token reduction) **Common operations:** - `analyze_change_impact(target, change_description, detail_level)` - Impact analysis **Usage in Cursor:** - "Analyze impact of changing TaskOrchestrator.HandleAsync" - "What would break if I modify this interface?" - "Show dependencies for this method" **Token Savings**: 85% (300 vs 2000 tokens) ### Context Optimization **Purpose**: Prepare optimized context for code generation (85% token reduction) **Common operations:** - `get_code_context(file_path, task_description, include_patterns)` - Optimized context **Usage in Cursor:** - "Prepare context for adding error handling to rag_agent.py" - "Get optimized context for this file before generating code" **Token Savings**: 85% (400 vs 2700 tokens) ### Code Generation **Purpose**: Generate code patches using local phi4 LLM (100% free) **Common operations:** - `generate_code_patch(file_path, change_description, validate_remote)` - Generate patch **Usage in Cursor:** - "Generate a patch to add timeout handling" - "Create a patch for error handling in this file" **Token Savings**: 100% (local generation, no remote tokens) ### Direct File Access (Use Sparingly) **Purpose**: Read full file content (expensive - use only when needed) **Common operations:** - `read_file_full(file_path)` - Read full file **Usage in Cursor:** - Only when optimized context insufficient - User explicitly requests full file - Verification after generation **Warning**: Expensive - sends full file to remote LLM ## Best Practices ### Token Efficiency **Community MCP Servers:** - Ask which files to read before reading them - Use git log with `--oneline` for quick history - Search before reading entire directories - Request summaries for large files **AIStack Intelligence Server:** - ✅ Use `search_code_semantic` instead of reading full files (90% savings) - ✅ Use `get_code_context` instead of sending full files (85% savings) - ✅ Use `generate_code_patch` for local generation (100% free) - ✅ Use `analyze_code_patterns` before generating code (95% savings) - ❌ Avoid `read_file_full` unless absolutely necessary ### Workflow 1. **Understand**: Use filesystem + git to understand context 2. **Plan**: Use sequential-thinking for complex changes 3. **Research**: Use brave-search for best practices 4. **Implement**: Generate code matching existing patterns 5. **Verify**: Use git to check changes ### Common Patterns **Understanding existing code:** - "Read src/TaskOrchestrator.cs and explain its purpose" - "Search for similar error handling patterns in the codebase" - "Show git history to understand why this was implemented this way" **Planning changes:** - "Use sequential-thinking to plan adding authentication" - "Break down this refactoring into safe, incremental steps" - "Analyze impact of changing this interface" **Researching solutions:** - "Search for best practices in async cancellation" - "Find examples of dependency injection in .NET 8" - "Research modern testing patterns for async code" **Using AIStack Intelligence:** - "Use aistack-intelligence to search for error handling patterns" - "Analyze the impact of changing this method" - "Generate a patch to add timeout handling" - "What patterns are used for dependency injection?"