Claudia

Built specifically for AI agents that need to coordinate multi-step work. Exposes hierarchical task management through MCP tools like task_start, task_finish, and task_handoff, letting Claude claim tasks atomically, track dependencies, and verify acceptance criteria before marking things complete. Each task gets 64KB of JSON context storage for agent memory between sessions. Includes sprint management, dependency graphs with cycle detection, and a kanban dashboard. The workflow enforces state transitions from pending through in_progress and verification to completed. Reach for this when you're using Claude for extended coding sessions and want it to manage its own todo list, remember what it was doing, or coordinate with other agent instances on the same codebase.