- Tool Descriptions: Rewritten all tool descriptions with token-aware guidance, expected output sizes, and usage recommendations to help AI assistants pick the right tool
get_content_details: Now returns summary + metadata by default (~200-500 tokens). NewincludeFullContentparameter to opt-in to full text retrievalsearch_content: WhenincludeFullContentis true, enriches results viabatchGetContentfor reliable full text. Renders full text in outputexplore_topic: Filters synthesized insights and frameworks for topic relevance (prevents off-topic results from matched sources). Better empty state with actionable suggestions- API URL: Changed default from
www.noverload.comtonoverload.com
- Replaced
anytypes with properRecord<string, unknown>in search implementation - Documentation URLs updated across README, INSTALL, and CHANGELOG
This beta release focuses on improving stability, reliability, and error handling based on early user feedback and testing.
- Enhanced error handling for expired tokens with clear user guidance
- Improved configuration validation with better error messages
- More robust API response error handling
- Clearer error messages for common configuration issues
- Configuration: Now exclusively uses environment variables for improved reliability and consistency
- Initialization: Streamlined process for better stability
- Token Validation: Enhanced flow with specific handling for authentication errors
- Error Reporting: More informative messages to help troubleshoot issues
- Configuration loading edge cases resolved
- TypeScript warnings cleaned up
- General stability improvements throughout the codebase
- Error message clarity enhanced
- Simplified configuration approach
- More predictable error handling
- Better alignment with MCP protocol best practices
- Improved debugging with clearer error messages
- BREAKING: Updated all API calls to use the new v2 endpoints
- Search endpoint:
/api/mcp/search→/api/mcp/v2/search - Content endpoint:
/api/mcp/content→/api/mcp/v2/content - Actions endpoint:
/api/mcp/actions→/api/mcp/v2/actions - Synthesis endpoint:
/api/mcp/synthesize→/api/mcp/v2/synthesis
- Search endpoint:
- Search: Now supports multiple modes (smart, hybrid, semantic, fulltext)
- Content Retrieval: Added token warnings and confirmation for large content
- Batch Operations: Enhanced batch content retrieval with enrichment options
- Response Handling: Adapted to new v2 response formats with better error handling
- Performance: Benefits from v2 API optimizations (caching, parallel processing)
- Maintains backward compatibility with existing MCP tool interfaces
- Response parsing automatically handles v2 format differences
- Error messages provide more context with error codes
- The MCP client now requires Noverload API v2 endpoints to be available
- Token warnings will appear for content over 10,000 tokens
- Large content (>25,000 tokens) requires explicit confirmation
- Initial MCP implementation with v1 API support
For more details on the v2 API, see the API Reference