Task tracking is the visible layer; `detect_failure_patterns` and `check_pattern_risk` are what make it more than a Kanban board with an API — the system flags an approach that has burned you before, using its own recorded history. `rollback_task` and the snapshot tools follow the same instinct. It is a large surface with a database dependency, so treat setup as a real step.
Orchestro sits between product intent and code. Write a user story and it decomposes it into technical tasks; from there it tracks dependencies and conflicts, enriches the implementation prompt, records what worked, and keeps a full history — with a live Kanban dashboard for people who do not read code.
- Tasks — `create_task`, `list_tasks`, `update_task`, `delete_task`, `get_execution_order` for a topological sort by dependency, `safe_delete_tasks_by_status`
- Execution flow — `prepare_task_for_execution` generates the codebase analysis prompt, `save_task_analysis` stores the result, `get_execution_prompt` returns enriched implementation context
- User stories — `decompose_story` turns a story into tasks with analysis on by default, plus `get_user_stories`, `get_tasks_by_user_story`, `get_user_story_health`
- Dependencies and conflicts — `save_dependencies`, `get_task_dependency_graph`, `get_resource_usage`, `get_task_conflicts` to catch two tasks touching the same files
- Knowledge and learning — `list_templates`, `list_patterns`, `list_learnings`, `render_template`, `get_relevant_knowledge`, `add_feedback`, `get_similar_learnings`, `get_top_patterns`, `get_trending_patterns`, `get_pattern_stats`, `detect_failure_patterns`, `check_pattern_risk`
- Project configuration — tech stack, guardian sub-agents including `sync_claude_code_agents` and `suggest_agents_for_task`, MCP tool registration with `suggest_tools_for_task`, guidelines and code patterns
- History — `get_task_history`, `get_status_history`, `get_decisions`, `record_decision`, `get_code_changes`, `get_guardian_interventions`, `get_task_snapshot` at a timestamp, and `rollback_task` to restore a previous state
Node.js 18+ and a Supabase project (the free tier is enough). Register `npx -y @khaoss85/orchestro@latest` in your Claude Code config with `DATABASE_URL` set to the Supabase connection string; `SUPABASE_URL` and `SUPABASE_SERVICE_KEY` come from the project's API settings and the service key must stay secret. Apply the schema with the migration script before first use. The real-time dashboard runs over Socket.io. Package version 2.1.0. MIT.
One command plus a key — npx @khaoss85/orchestro@latest, then supply credentials