Context Optimizer MCP Server

# Changelog All notable changes to the Context Optimizer MCP Server project will be documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). ## [1.0.7] - 2025-08-30 ### Changed - **BREAKING**: Refactored Exa Research API integration to use only latest SDK methods - **Research Tools**: Now uses only `research.create`, `research.get`, and `research.pollUntilFinished` - **API Compatibility**: Dropped backward compatibility for legacy methods (`createTask`, `getTask`, `pollTask`) - **Polling Logic**: Simplified polling fallback mechanism for better reliability ### Removed - **Legacy API Support**: Removed fallback support for older Exa Research API methods ## [1.0.6] - 2025-08-30 ### Changed - **Version Bump**: Preparation release for upcoming changes ## [1.0.5] - 2025-08-30 ### Added - **CI/CD**: PR validation workflow with automated build, test, and version validation - **Documentation**: Manual testing framework with comprehensive workflow-based protocols - **Test Data**: Unicode and large file test cases for better coverage ### Fixed - **Cross-Platform**: Fixed Windows-specific test failures on Linux/macOS platforms - **Test Reliability**: Improved platform-specific test execution and logging - **exa-js Compatibility**: Added full support for exa-js 1.9+ API changes with intelligent fallbacks ### Changed - **Dependency Management**: Pinned exa-js to version 1.9.1 via npm overrides to prevent version drift - **Error Handling**: Enhanced error diagnostics with installed package version information - **npm Publishing**: Improved error handling - now fails on duplicate versions instead of skipping - **Testing**: Reduced test output noise by changing log level from 'info' to 'error' during tests ### Security - **API Detection**: Replaced fragile require() calls with robust API detection for bundled environments ## [1.0.4] - 2025-08-30 ### Added - **npm Trusted Publishing**: Implemented OIDC authentication for enhanced security - **Workflow Security**: Added optional npm-publish environment for additional protection - **Version Validation**: Automated checks to ensure package version is greater than published version ### Changed - **CI/CD**: Upgraded npm and improved publishing workflow reliability - **Authentication**: Removed dependency on NPM_TOKEN secrets in favor of OIDC - **Publishing**: Enhanced publish step messaging and error handling ### Removed - **Token Authentication**: Eliminated NPM_TOKEN secret requirement for publishing ### Security - **OIDC Integration**: Enhanced security with OpenID Connect for npm publishing - **Secret Management**: Reduced attack surface by eliminating token-based authentication ## [1.0.3] - 2025-08-27 ### Fixed - **Exa Research Integration:** Updated to `exa-js@^1.5.13`, switched to default import, and added compatibility with `research.pollTask`. Resolved runtime error `createTask is not a function` for `researchTopic` and `deepResearch` tools. ### Changed - **VS Code MCP Config Guidance:** Documented how to point VS Code to the local `dist/server.js` for development instead of the globally installed package. ### Added - **npm Publishing Workflow:** Added automated workflow to publish packages to npm registry ## [1.0.2] - 2025-08-08 ### Fixed - **CLI Version Display**: Fixed `--version` and `--help` commands to show correct version number dynamically from package.json - **Server Configuration**: MCP server now reports correct version in protocol metadata ## [1.0.1] - 2025-08-08 ### Added - **CLI Argument Support**: Added `--version` and `--help` command line arguments - **Graceful Exit**: Server now exits properly for CLI commands instead of hanging ## [1.0.0] - 2025-08-08 ### Added - **Core MCP Tools**: File analysis (`askAboutFile`), terminal execution (`runAndExtract`), follow-up questions (`askFollowUp`) - **Research Tools**: Quick research (`researchTopic`) and deep research (`deepResearch`) using Exa.ai - **Multi-LLM Support**: Google Gemini, Anthropic Claude, and OpenAI integration - **Security Layer**: Path validation, command filtering, session management - **Environment Variable Configuration**: All settings via `CONTEXT_OPT_*` environment variables - **MCP Protocol Compliance**: Proper JSON-RPC communication over stdio - **Cross-Platform Support**: Windows, macOS, and Linux compatibility - **Comprehensive Testing**: Unit tests, integration tests, security validation - **Documentation**: Setup guides for VS Code, Claude Desktop, and Cursor AI ### Fixed - **MCP Protocol Issues**: Fixed stdout/stderr logging conflicts causing "Failed to parse message" warnings - **Windows Path Validation**: Fixed case sensitivity issues in path validation - **Environment Variable Loading**: Consistent `CONTEXT_OPT_` prefixed variables across all components - **Test Suite**: All 84 tests passing with conditional LLM integration testing ### Changed - **Logging**: Default log level changed from 'info' to 'warn' for cleaner output - **Configuration**: Moved from JSON config files to environment variables only - **Error Handling**: Enhanced MCP-compliant error responses with proper content structure - **Documentation Updates**: Enhanced README.md with detailed tool summaries including specific use cases and capabilities - **Usage Guide Improvements**: - Merged required and optional environment variables into single unified section - Removed temporary/session-only environment variable options for cleaner setup - Updated AI Assistant Setup with accurate MCP configuration based on latest 2024-2025 specifications - Verified VS Code, Claude Desktop, and Cursor AI setup procedures with proper file locations and JSON schemas