Orchestro is an intelligent orchestration system for AI-assisted development, positioned as "your AI development conductor." It aims to solve problems such as product managers finding it difficult to track development progress, developers frequently switching contexts and dependencies, loss of knowledge between Claude Code sessions, and the lack of a single source of truth, by orchestrating the entire development lifecycle through task decomposition, dependency tracking, pattern learning, and real-time progress visualization.
For product managers and leads: write user stories in natural language, and AI automatically generates technical tasks; provides a real-time updated Kanban view; manage development without technical knowledge; automatically flagged risks explained in plain language; fully transparent display of what is being built, when, and why.
For developers: AI analyzes the codebase and provides implementation suggestions; displays task dependencies as a visual dependency graph; the system learns coding patterns from successes and failures; detects whether multiple tasks touch the same files to prevent conflicts; preserves context between sessions.
For Claude Code: provides 60 MCP tools; structured workflow (prepare → analyze → implement → learn); context-aware implementation prompts; a persistent knowledge base of templates, patterns, and experiential knowledge; real-time dashboard updates via Socket.io; complete decision and change timeline; rollback capability; exportable Markdown reports for stakeholders.
Main capabilities include: user story decomposition (decompose_story, with autoAnalyze=true by default to automatically prepare analysis prompts), task management (create/list/update/delete, filter by status or tag, topological sorting of execution order), task execution and analysis (generate codebase analysis prompts, save analysis results, obtain enhanced execution prompts), dependencies and conflicts (save dependencies, dependency graph, resource occupancy, conflict detection), knowledge and templates (templates, patterns, experience lists and rendering, context-related suggestions), feedback and learning (record successes/failures/improvements, similar experiences, high-frequency/trend patterns, pattern statistics, failure mode detection, pattern risk checks), project configuration (tech stack, sub-agents/guardians, MCP tools, coding guidelines and code patterns), task history and events (complete event timeline, status history, decision records, guardian interventions, code changes, rollback, snapshots, statistics).
Dashboard features: the Kanban supports drag-and-drop user stories, instant progress viewing, plain-language risk indicators, and report export; the developer view provides technical task details, dependency indicators, code complexity badges, and direct file links; the task details page is divided into overview (for PMs), history (audit trailtraces and rollback), and dependencies (dependency graph, resource impact, risk assessment, conflict detection).
Installation methods include: adding from the official MCP registry (recommended), one-click installation (automatically downloads configuration, applies database migrations to Supabase, configures Claude Code, starts the dashboard and verifies), manual installation (create a Supabase project, copy the connection string, apply the database schema, configure credentials, requires setting the SUPABASE_URL, SUPABASE_SERVICE_KEY, and ANTHROPIC_API_KEY environment variables), adding to an existing Claude Code project, or installing as a Claude Code plugin (includes MCP server configuration, 5 guardian agents: database, API, architecture, test-maintainer, production-ready, and complete documentation).
Architecturally, the backend uses @modelcontextprotocol/sdk, Supabase (PostgreSQL), and Socket.io for real-time communication; the frontend uses Next.js 14 (App Router), React 18, TypeScript, TailwindCSS + shadcn/ui, React Flow (graphics), and react-markdown. The core database tables include projects, tasks, and task_dependencies; the knowledge category includes learnings, patterns, templates, and pattern_frequency; the resource category includes resource_nodes, resource_edges, code_entities, and code_dependencies; the system category includes event_queue, file_history, and codebase_analysis, and uses JSONB metadata, GIN indexes, and row-level security (RLS).
In terms of performance and scale: query speed under GIN indexes is <10ms; the real-time polling interval is 1 second; processed events are automatically cleaned up after 24 hours; 100+ tasks have been tested; analysis is non-blocking (delegated to Claude); supports multiple PMs and multiple developers. For security, it uses environment variables to store secrets, Supabase RLS, a complete audit trail,event processing to prevent duplicate operations, data local-first (stored in your own Supabase instance), and supports export and deletion to comply with GDPR.
Typical example: a product manager enters the "e-commerce checkout flow" story, and Orchestro automatically breaks it down into 6 tasks: database design, payment service integration, checkout API, checkout UI component, email notification, and order confirmation flow, and automatically maps dependencies (total estimate 18 hours). Developers execute according to the workflow, and the kanban board updates progress and blocked status in real time.
The roadmap shows: Phase 1 core orchestration (60 MCP tools, real-time kanban, AI story breakdown, pattern learning and failure detection, dependency and conflict detection, task metadata, audit trail, project configuration, Claude Code agent synchronization, AI agent and tool suggestions) is complete; Phase 2 PM enablement (non-technical views, story templates, report export, notifications, plain-language risk explanations) is in progress; Phase 3 team intelligence (multi-team workspaces, cross-project pattern sharing, velocity tracking and estimation, automatic assignment, Slack/Teams integration) and Phase 4 advanced AI (LangGraph automatic orchestration, predictive risk detection, automatic conflict resolution, code review automation, documentation generation) are planned. Documentation includes QUICK_START.md, PM_GUIDE.md (coming soon), DEV_GUIDE.md (coming soon), EXAMPLES.md, WORKFLOW.md, ARCHITECTURE.md, and API.md. The current version v2.1.0 (2025-10-10) has been published to the official MCP registry and npm (@khaoss85/orchestro), and all 60 tools have been tested and verified (96.7% success rate). Specific pricing and billing models are not yet available.