Tasks MCP Server

Work in mcp-tasks is organized around task files rather than direct AI edits to documents. tasks_setup initializes a source, creates a missing file, and returns the identifier used for later calls. Markdown, JSON, and YAML files can coexist across multiple sources, with changes saved automatically. tasks_add takes batches of task text and a destination status; tasks_update moves existing tasks by ID and can adjust their position. To locate work, tasks_search supports status filters, ID lookup, and case-insensitive text matching with OR logic.

tasks_summary brings together counts for each status and the current work-in-progress tasks. Status names are configurable, and automatic work-in-progress handling can move other active tasks back to To Do or promote the first To Do item when none is active. The CLI offers the same operations with JSON output, including a way to add tasks from a separate terminal while AI work continues. Dedicated task files are important: Markdown rewrites preserve recognized task sections, so unrelated notes or documentation may be lost.