Codebase MCP Server

# Phase 02: Remove Non-Search MCP Tools ## Problem Statement The codebase-mcp server currently exposes 16 MCP tools, but we only need 2 (index_repository and search_code) for semantic code search. The 14 non-search tools and their supporting code add ~2,700 unnecessary lines of code, increasing maintenance burden and violating the constitutional principle of "Simplicity Over Features". After Phase 01 used Alembic migration 002 to drop 9 database tables (archived_work_items, work_item_deployment_links, vendor_deployment_links, work_item_dependencies, future_enhancements, project_configuration, deployment_events, vendor_extractors, tasks), the code referencing those tables must be removed to avoid errors. ## User Stories ### As a Maintainer I want non-search MCP tools removed from the server, so that I only have to maintain code relevant to semantic search functionality. ### As a User I want a focused MCP server with clear purpose (semantic search only), so that I understand exactly what capabilities it provides without confusion from unrelated features. ### As a Developer I want supporting code for removed features deleted (database operations, Pydantic models, tests), so that the codebase is clean and doesn't reference non-existent database tables. ### As a System Administrator I want the codebase reduced by 60% (~4,500 to ~1,800 lines), so that deployments are faster, security audits are easier, and the attack surface is smaller. ## Success Criteria ### Tools Removed - Only 2 MCP tools remain registered in server: index_repository, search_code - 14 tools removed: create_work_item, update_work_item, query_work_item, list_work_items, create_task, get_task, list_tasks, update_task, create_vendor, query_vendor_status, update_vendor_status, record_deployment, get_project_configuration, update_project_configuration ### Code Deleted - 14 tool implementation files deleted from `src/mcp/tools/` - Database operation modules for removed features deleted - Pydantic models for removed features deleted - Server registration code updated to only register 2 tools - All imports for deleted code removed ### Tests Cleaned Up - Test files for removed features deleted - Only search-related tests remain in `tests/` - All remaining tests pass - Type checking passes: `mypy --strict src/` ### Code Reduction Achieved - Total lines of code reduced from ~4,500 to ~1,800 (60% reduction) - Only search-related code remains in src/ directory ### Server Functionality - Server starts successfully with only 2 tools - No import errors or references to deleted code - MCP protocol compliance maintained ## Constraints ### Preserve Search Functionality (NON-NEGOTIABLE) - index_repository tool must remain fully functional - search_code tool must remain fully functional - No changes to search algorithm or indexing logic - All search-related tests must continue to pass ### Type Safety (NON-NEGOTIABLE) - `mypy --strict` must pass after all deletions - No type: ignore comments added (or justified if necessary) - Pydantic models for remaining features must remain valid ### Incremental Deletion Strategy - Delete in 4 sub-phases (tool files, database operations, server registration, tests) - One git commit per sub-phase for clear history - Verify compilation after each deletion step ### No Database Changes - This phase only removes code, not database schema - Database changes were completed in Phase 01 ## Out of Scope ### Not Included in This Phase - Multi-project feature implementation (that's Phase 03) - Connection pool implementation (that's Phase 04) - Documentation updates (that's Phase 05) - Performance optimization (that's Phase 06) ### Explicitly NOT Doing - Modifying search or indexing algorithms - Changing remaining tool interfaces - Database schema modifications - Configuration changes ## Business Value ### Reduced Maintenance Burden With 60% less code, there are fewer files to maintain, fewer dependencies to manage, fewer potential bugs, and simpler onboarding for new developers. ### Improved Security Posture Smaller codebase means smaller attack surface. Fewer dependencies means fewer potential vulnerabilities. Security audits become faster and more thorough. ### Faster Deployments Smaller codebase means faster Docker builds, faster deployments, faster startup time, and reduced disk space usage. ### Clearer Purpose Removing non-search features makes the purpose crystal clear: this is a semantic code search server. No confusion about feature scope or mission creep. ### Constitutional Compliance Aligns with Constitutional Principle #1: "Simplicity Over Features - Focus exclusively on semantic code search". ## Additional Context This phase corresponds to Phases 3-6 from FINAL-IMPLEMENTATION-PLAN.md. It should take 8-12 hours to complete and depends on Phase 01 (Alembic migration 002 applied: 9 tables dropped - archived_work_items, work_item_deployment_links, vendor_deployment_links, work_item_dependencies, future_enhancements, project_configuration, deployment_events, vendor_extractors, tasks - and project_id columns added to repositories and code_chunks). The deletion should be done incrementally in 4 sub-phases: 1. Delete tool implementation files 2. Delete database operation modules 3. Update server registration 4. Delete tests Each sub-phase should have its own git commit for clear history and easy rollback if needed.