mcp-server-tree-sitter

# MCP Tree-sitter Server Roadmap This document outlines the planned improvements and future features for the MCP Tree-sitter Server project. CRITICAL: When a task is done, update this document to mark it done. However, you must ensure it is done for all files/subjects present in the repo. DO NOT mark a task done simply because a subset of the targeted files/subjects have been handled. Mark it [WIP] in that case. ## Short-term Goals ### Code Quality - ✅ Fix linting issues identified by ruff - ✅ Improve exception handling using proper `from` clause - ✅ Remove unused variables and improve code organization - ✅ Implement TreeCursor API support with proper type handling - ✅ Add incremental parsing support - ✅ Add MCP Progress Reporting - ✅ Add Server Capabilities Declaration - [ ] Add mcp server start flag(s) for enabling (allow list approach) and disabling (block list approach) a list of features. Only one approach may be applied at a time. The default should be minimal allowed, for now. Add meta features such as stable, wip, advanced, basic - ✅ Add mcp server start flag(s) for ensuring language packs are installed - Resolved by tree-sitter-language-pack integration - [ ] Add mcp server start flag(s) for ensuring project is configured beforehand. - [ ] Achieve 100% type hinting coverage (and ensure this is enforced by our linting) - [ ] Improve docstring coverage and quality (Don't thrash on updating docs that are already good) (HOLD pending other work) - [ ] Split files until the longest .py file is less than 500 lines (unless that breaks functionality, in which case do not) ### Testing - ✅ Create and maintain tests for AST functionality, query execution, and symbol extraction - 🔄 [WIP] Create additional tests for context utilities, incremental parsing, and cursor traversal - [ ] Increase unit test coverage to 100% and begin enforcing that in pre-commit and CI - [ ] Add integration tests for MCP server functionality (HOLD pending other work) - [ ] Create automated testing workflow with GitHub Actions (unit, integration, static, etc) (HOLD pending other work) ### Documentation (HOLD) - ✅ Create CONTRIBUTING.md with developer guidelines - 🔄 [WIP] Create a docs/user-guide.md with more examples and clearer installation instructions. Link to it from README.md - [ ] Add detailed API documentation in docs/api-guide.md - 🔄 [WIP] Create usage tutorials and examples -- focus only on Claude Desktop for now. ## Medium-term Goals (HOLD) ### Feature Improvements - ✅ Add support for more tree-sitter languages by implementing https://github.com/Goldziher/tree-sitter-language-pack/ - ✅ Add support for query execution with proper result handling - [ ] Improve query building tools with more sophisticated matching options (HOLD because we could cripple the codebase with complexity) - [ ] Implement more advanced code analysis metrics (HOLD because we could cripple the codebase with complexity) - [ ] Enhance caching system with better invalidation strategy (HOLD because we could cripple the codebase with complexity) ### User Experience - [ ] Create a web-based UI for visualizing ASTs and running queries (HOLD because Claude's experience is more important) - [ ] Add CLI commands for common operations (HOLD because Claude runs commands by a different channel) - [✅] Implement progress reporting for long-running operations - [ ] Add configuration presets for different use cases (HOLD because we could cripple the codebase with complexity) ### Security - [ ] Add comprehensive input validation (HOLD because we could cripple the codebase with complexity) - [ ] Implement access control for multi-user environments (HOLD because we could cripple the codebase with complexity) - [ ] Add sandbox mode for running untrusted queries (HOLD because we could cripple the codebase with complexity) ## Long-term Goals (HOLD) ### Advanced Features - [ ] Implement semantic analysis capabilities (HOLD because we need stability first) - [ ] Add code transformation tools (HOLD because we need stability first) - [ ] Support cross-language analysis (HOLD because we need stability first) ### Integration - [ ] Create plugins for popular IDEs (VS Code, IntelliJ) (HOLD because we need stability first) - [ ] Implement integration with CI/CD pipelines (HOLD because we need stability first) - [ ] Add support for other LLM frameworks beyond MCP (HOLD because we need stability first) ### Performance - [ ] Optimize for large codebases (> 1M LOC) (HOLD because we need stability first) - [ ] Implement distributed analysis for very large projects (HOLD because we need stability first) - [ ] Add streaming responses for large result sets (HOLD because we need stability first) ## Completed Implementations ### MCP Context Handling - Added `utils/context/mcp_context.py` with progress tracking capabilities - Implemented `MCPContext` class with progress reporting - Created `ProgressScope` for structured operation tracking - Added context information passing to analysis tools ### TreeCursor API Support - Enhanced `utils/tree_sitter_types.py` with TreeCursor protocol - Added efficient cursor-based tree traversal in `utils/tree_sitter_helpers.py` - Implemented collector pattern using cursors to efficiently find nodes ### Incremental Parsing - Added support for tree editing in `utils/tree_sitter_helpers.py` - Enhanced cache to track tree modifications in `cache/parser_cache.py` - Implemented changed_ranges detection for optimization ### Server Capabilities Declaration - Created `capabilities/server_capabilities.py` for capability declaration - Implemented required MCP server capabilities - Added support for completion suggestions - Added structured logging integration ## Features and Ideas Below are some ideas and feature requests being considered: 1. **Semantic Diff**: Show semantic differences between code versions rather than just text diffs (HOLD because we need stability first) 2. **Code Quality Metrics**: Integrate with code quality metrics and linters (HOLD because we need stability first) 3. **Interactive Query Builder**: Visual tool to build and test tree-sitter queries (HOLD because we need stability first) 4. **Code Completion**: Use tree-sitter for more intelligent code completion suggestions (HOLD because we need stability first) 5. **Visualization Export**: Export AST visualizations to various formats (SVG, PNG, etc.) (HOLD because we need stability first)