Tasks

Stores and retrieves tasks from Markdown, JSON, and YAML files with automatic format detection based on file extension and content parsing. The system maintains a unified in-memory task model while delegating serialization/deserialization to format-specific handlers, enabling seamless switching between storage formats without data loss or schema migration.

Provides structured filtering and full-text search capabilities designed to reduce LLM context window consumption by returning only relevant tasks. Uses indexed search patterns and filter predicates to avoid sending entire task databases to the LLM, with support for filtering by status, priority, tags, and date ranges while maintaining O(n) or better performance characteristics.

Exposes task management operations as MCP (Model Context Protocol) tools with JSON schema definitions, enabling LLMs to discover, understand, and invoke task operations through standardized function-calling interfaces. Each operation (create, read, update, delete, search) is registered as a callable tool with input/output schemas that guide LLM behavior and validate arguments before execution.

Supports flexible task organization through multi-level tagging, custom metadata fields, and status tracking without enforcing rigid hierarchies. Tasks can be tagged with multiple labels, assigned custom properties, and tracked through configurable status workflows, enabling diverse organizational patterns (GTD, Kanban, priority-based) without schema changes.

Implements create, read, update, and delete operations optimized for LLM agent invocation, with minimal argument complexity and clear success/failure semantics. Each operation is designed to be callable with minimal context and returns concise results to avoid wasting LLM tokens on verbose responses, using operation-specific schemas that guide LLM behavior toward efficient calls.

Reduces tool confusion by providing a minimal, well-defined set of task operations with clear, non-overlapping responsibilities and unambiguous naming. Each tool has a single, obvious purpose (e.g., 'create_task' vs 'update_task' vs 'search_tasks'), with schemas that prevent the LLM from misusing operations or confusing similar tools, and documentation that guides correct usage patterns.