Your AI Development Conductor - From Product Vision to Production Code
Transform product ideas into reality with an intelligent orchestration system that bridges Product Managers, Developers, and AI. Orchestro conducts the entire development symphony: task decomposition, dependency tracking, pattern learning, and real-time progress visualization.
The Problem:
- Product Managers lose track of development progress
- Developers struggle with context switching and dependencies
- Knowledge is lost between Claude Code sessions
- No single source of truth for what's being built
The Solution: Orchestro orchestrates the entire development lifecycle:
- ๐ For PMs: Visual Kanban board, user story decomposition, progress tracking
- ๐จโ๐ป For Developers: AI-powered task analysis, dependency graphs, pattern learning
- ๐ค For Claude Code: Structured workflows, enriched context, knowledge retention
- ๐ For Everyone: Real-time dashboard, transparent progress, complete audit trail
Think Trello ร Jira ร AI - but designed specifically for AI-assisted development.
- User Story Decomposition - Write a story, AI creates technical tasks automatically
- Visual Progress Board - Kanban view with real-time updates
- No Technical Knowledge Required - Manage development without coding
- Complete Transparency - See exactly what's being built, when, and why
- Risk Awareness - Auto-flagged risks with plain English explanations
- Intelligent Task Analysis - AI analyzes codebase and suggests implementation
- Dependency Tracking - Visual graphs show what depends on what
- Pattern Learning - System learns from successes and failures
- Conflict Prevention - Detects when tasks touch the same files
- Context Retention - Never lose context between sessions
- 60 MCP Tools - Complete toolkit for orchestrated development
- Structured Workflows - prepare โ analyze โ implement โ learn
- Enriched Prompts - Context-aware implementation guidance
- Knowledge Base - Templates, patterns, learnings persist forever
- Real-Time Dashboard - Live updates via Socket.io
- Complete History - Timeline of all decisions and changes
- Rollback Capability - Undo mistakes safely
- Export Everything - Markdown reports for stakeholders
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ PRODUCT MANAGER โ
โ "User should login with email/password" โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ
โโโโโโโโโโโโโโโโโโโโโโโโ
โ ORCHESTRO AI โ
โ Decomposes Story โ
โโโโโโโโโโโโโโโโโโโโโโโโ
โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ 7 Technical Tasks Created โ
โ โข Database schema โ
โ โข Authentication service โ
โ โข API endpoints โ
โ โข Frontend components โ
โ โข State management โ
โ (with dependencies automatically) โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ
โโโโโโโโโโโโโโโโโโโโโโโโ
โ DEVELOPER/CLAUDE โ
โ Implements Tasks โ
โโโโโโโโโโโโโโโโโโโโโโโโ
โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ PM SEES PROGRESS โ
โ โข Kanban updates in real-time โ
โ โข Risks flagged automatically โ
โ โข Dependencies visualized โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
npx @orchestro/init
That's it! The installer will:
- โ Download and setup Orchestro
- โ Apply database migrations to Supabase
- โ Configure Claude Code automatically
- โ Setup Supabase connection
- โ Start the dashboard
- โ Verify everything works
Interactive prompts:
๐ญ Orchestro Setup Wizard
? Supabase connection string: โโโโโโโโ
? Project name: My Project
? Install location: ~/orchestro
โ๏ธ Setting up...
โ Orchestro installed
โ Claude Code configured
โ Database ready
๐ Done! Restart Claude Code and ask:
"Show me orchestro tools"
# Node.js 18+
node --version
# Supabase account (free tier works great)
# Sign up at https://supabase.com
Create your Supabase project:
- Go to https://supabase.com and create a new project
- Wait for the database to be provisioned (~2 minutes)
- Go to Settings โ Database and copy the Connection String (Transaction mode)
Apply database schema:
# Clone this repo first
git clone https://github.com/yourusername/orchestro.git
cd orchestro
# Install dependencies
npm install
# Set your Supabase connection string
export DATABASE_URL="your-supabase-connection-string"
# Apply all migrations to create the schema
npm run migrate
Verify database setup:
# The migrate script will show you all tables created:
# You should see:
# โ
Running migration: code_entities
# โ
Running migration: add_tasks_metadata
# โ
Running migration: fix_status_transition_trigger
# โ
Running migration: event_queue
# โ
Running migration: auto_update_user_story_status
# โ
Running migration: add_task_metadata_fields
# โ
Running migration: add_pattern_frequency_tracking
# Or verify manually via Supabase dashboard:
# Go to Database โ Tables and check all tables are created
Get your credentials:
# From Supabase Dashboard:
# 1. DATABASE_URL (for migrations & MCP server)
# Settings โ Database โ Connection String โ Transaction mode
# Example: postgresql://postgres:[password]@db.[project].supabase.co:5432/postgres
# 2. SUPABASE_URL (for API calls)
# Settings โ API โ Project URL
# Example: https://[project].supabase.co
# 3. SUPABASE_SERVICE_KEY (for admin operations - keep secret!)
# Settings โ API โ service_role key
# Example: eyJhbG...
# Run interactive setup
npm run setup
# Or manual configuration:
cat > .env << EOF
DATABASE_URL=your-supabase-connection-string
SUPABASE_URL=your-supabase-url
SUPABASE_SERVICE_KEY=your-service-key
EOF
# Auto-configure (recommended)
npm run configure-claude
# Or manually edit:
open ~/Library/Application\ Support/Claude/claude_desktop_config.json
# Add:
{
"mcpServers": {
"orchestro": {
"command": "node",
"args": ["/absolute/path/to/orchestro/dist/server.js"],
"env": {
"DATABASE_URL": "your-connection-string"
}
}
}
}
npm run dashboard
# ๐ Opens http://localhost:3000
# Restart Claude Code, then ask:
"Show me all orchestro tools"
# You should see 60 tools! ๐ญ
Already have a Claude Code project? Add Orchestro:
# In your project directory
npx @orchestro/add
# Or via Claude Code config:
claude mcp add orchestro
See Integration Guide for existing project setup.
Scenario: New feature request from stakeholder
1. Write user story in dashboard:
"User should be able to export report as PDF"
2. Click "Decompose with AI"
โ Orchestro creates 5 technical tasks with dependencies
3. Monitor Kanban board:
โ See real-time progress as Claude implements
โ Risks flagged automatically (e.g., "PDF library size impact")
โ Hover over task for technical details
4. Review & Accept:
โ See code diffs in plain English
โ Rollback if needed
โ Export timeline for stakeholder report
Scenario: Implementing complex feature
1. Pick task from Kanban board
2. Ask Claude:
"Prepare task [task-id] for execution"
โ Orchestro analyzes codebase
โ Shows: files to modify, dependencies, risks
3. Get enriched context:
โ Past similar implementations
โ Relevant patterns (with success rates!)
โ Risk mitigation strategies
4. Implement with confidence:
โ Conflict detection warns if other tasks touch same files
โ Pattern learning suggests best approaches
โ Complete history for rollback safety
Scenario: Cross-functional collaboration
PM writes story โ AI decomposes โ Dev implements โ All see progress
โข PM: Non-technical Kanban view
โข Dev: Technical dependency graph
โข Claude: Enriched implementation context
โข Everyone: Real-time updates, complete transparency
get_project_info
- Project metadata and statusget_project_configuration
- Complete project configurationinitialize_project_configuration
- Setup default tools and guardians
create_task
- Create with assignee, priority, tags, categorylist_tasks
- Filter by status/category/tagsupdate_task
- Modify any field with validationdelete_task
- Safe deletion with dependency checksget_task_context
- Full context with dependencies (deprecated, use prepare_task_for_execution)get_execution_order
- Topological sort by dependenciessafe_delete_tasks_by_status
- Bulk delete with safety checks
prepare_task_for_execution
- Generate codebase analysis promptsave_task_analysis
- Store analysis resultsget_execution_prompt
- Enriched implementation context
decompose_story
- AI-powered story โ tasks decomposition with automatic analysis (autoAnalyze=true default)get_user_stories
- List all user storiesget_tasks_by_user_story
- Get all child tasksget_user_story_health
- Monitor story completion status
save_dependencies
- Record task resource dependenciesget_task_dependency_graph
- Visualize dependency graphget_resource_usage
- Find tasks using a resourceget_task_conflicts
- Detect conflicting resource usage
list_templates
- Available prompt/code templateslist_patterns
- Coding patterns librarylist_learnings
- Past experience recordsrender_template
- Generate from template with variablesget_relevant_knowledge
- Context-aware suggestions
add_feedback
- Record success/failure/improvementget_similar_learnings
- Find related experiencesget_top_patterns
- Most frequently used patternsget_trending_patterns
- Recently popular patternsget_pattern_stats
- Detailed pattern metricsdetect_failure_patterns
- Auto-detect risky approachescheck_pattern_risk
- Risk assessment before using pattern
Tech Stack:
add_tech_stack
- Add framework/libraryupdate_tech_stack
- Update version/configremove_tech_stack
- Remove technology
Sub-Agents (Guardians):
add_sub_agent
- Register guardian agentupdate_sub_agent
- Modify agent configsync_claude_code_agents
- Sync from .claude/agents/read_claude_code_agents
- Read agent filessuggest_agents_for_task
- AI-powered agent recommendationsupdate_agent_prompt_templates
- Update prompt templates
MCP Tools Management:
add_mcp_tool
- Register MCP toolupdate_mcp_tool
- Update tool configsuggest_tools_for_task
- AI-powered tool recommendations
Guidelines & Patterns:
add_guideline
- Add coding guidelineadd_code_pattern
- Add reusable pattern
get_task_history
- Complete event timelineget_status_history
- Status transition logget_decisions
- Decision recordsget_guardian_interventions
- Guardian activity logget_code_changes
- Code modification historyrecord_decision
- Log a decision with rationalerecord_code_change
- Log code modificationsrecord_guardian_intervention
- Log guardian actionrecord_status_transition
- Log status changeget_iteration_count
- Count task iterationsget_task_snapshot
- Task state at timestamprollback_task
- Restore previous stateget_task_stats
- Aggregate statistics
PM View:
- Drag & drop user stories
- See progress at a glance
- Risk indicators in plain English
- Export reports for stakeholders
Developer View:
- Technical task details
- Dependency indicators
- Code complexity badges
- Direct links to files
Tab: Overview (PM-friendly)
- User story description
- Technical requirements
- Assignee & priority
- Dependencies explained
Tab: History (Audit trail)
- Complete event timeline
- Decision records with rationale
- Code changes (with diffs)
- Rollback capability
Tab: Dependencies (Developer focus)
- Visual dependency graph
- Resource impact analysis
- Risk assessment
- Conflict detection
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ PRODUCT MANAGER โ
โ โข Writes user stories โ
โ โข Monitors Kanban board โ
โ โข Reviews progress โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ (Dashboard)
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ ORCHESTRO DASHBOARD (Next.js) โ
โ โข Kanban board with real-time updates โ
โ โข Dependency graphs โ
โ โข Progress visualization โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ โ (Socket.io)
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ SUPABASE (Data Layer) โ
โ โข Tasks, dependencies, resources โ
โ โข Event queue & real-time sync โ
โ โข Knowledge base & pattern tracking โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ โ (PostgreSQL)
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ ORCHESTRO MCP SERVER (Conductor) โ
โ โข 27 tools for task orchestration โ
โ โข Pattern learning & risk detection โ
โ โข AI story decomposition โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ โ (MCP Protocol)
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ CLAUDE CODE (Developer + AI) โ
โ โข Analyzes codebase โ
โ โข Implements features โ
โ โข Records decisions โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
PM writes in dashboard:
"Customer should complete purchase with
credit card payment and email confirmation"
Orchestro decomposes (AI-powered):
- โ Design checkout database schema (2h) - No dependencies
- โ Implement payment service integration (4h) - Depends on: #1
- โ Create checkout API endpoints (3h) - Depends on: #2
- โ Build checkout UI components (4h) - Depends on: #3
- โ Add email notification service (2h) - Depends on: #3
- โ Implement order confirmation flow (3h) - Depends on: #4, #5
Total: 18 hours, 6 tasks, dependencies mapped automatically
Developer flow (with autoAnalyze=true):
// 1. Decompose story (auto-analyzes tasks)
decompose_story("Customer checkout with payment")
// โ Creates 6 tasks
// โ Auto-generates analysis prompts for tasks without dependencies
// โ Returns analysisPrompts[] ready to use
// 2. Claude reviews analysis prompts
// Prompts include: files to check, patterns to find, risks to identify
// 3. Claude analyzes codebase using the prompts
// Finds: existing payment tables, similar schemas
// Risks: None (new table)
// 4. Save analysis results
save_task_analysis({
taskId: "task-1-id",
filesToCreate: ["migrations/002_checkout.sql"],
dependencies: [{type: "file", name: "001_orders.sql", action: "uses"}],
risks: []
})
// 5. Get enriched context
get_execution_prompt("task-1-id")
// โ Returns: related code, patterns, guidelines
// 6. Implement!
// Claude writes migration, runs tests
// 7. Record learning
add_feedback({
pattern: "e-commerce checkout schema",
type: "success",
feedback: "Stripe integration smooth"
})
Key improvement: Step 1 now auto-prepares analysis, reducing manual workflow steps!
PM sees:
- โ Task 1 โ Done (real-time update)
- ๐ก Task 2 โ In Progress (Claude working)
- โณ Tasks 3-6 โ Blocked (waiting for dependencies)
- ๐ Progress: 17% (1/6 tasks done)
// Scenario: Regex parsing keeps failing
// Attempt 1
add_feedback({
pattern: "regex pattern matching",
type: "failure",
feedback: "Unescaped metacharacters broke parser"
})
// Attempt 2
add_feedback({
pattern: "regex pattern matching",
type: "failure",
feedback: "Special chars not sanitized"
})
// Attempt 3
add_feedback({
pattern: "regex pattern matching",
type: "success",
feedback: "Finally worked after sanitizing"
})
// Now Orchestro knows...
detect_failure_patterns()
// ๐จ Returns:
// {
// pattern: "regex pattern matching",
// failure_rate: 66.67%,
// risk_level: "medium",
// recommendation: "โก Review sanitization first!"
// }
// Next time, before using regex:
check_pattern_risk("regex pattern matching")
// โ ๏ธ Warning: "67% failure rate (2/3).
// Common issue: Unescaped metacharacters.
// Mitigation: Use sanitization helper first."
Result: Future regex tasks complete faster with fewer errors!
Backend (MCP Server)
- TypeScript 5.0
- @modelcontextprotocol/sdk
- Supabase (PostgreSQL)
- Socket.io for real-time
Frontend (Dashboard)
- Next.js 14 (App Router)
- React 18 + TypeScript
- TailwindCSS + shadcn/ui
- React Flow (graphs)
- react-markdown (rendering)
Database (Supabase/PostgreSQL)
- Core: projects, tasks, task_dependencies
- Knowledge: learnings, patterns, templates, pattern_frequency
- Resources: resource_nodes, resource_edges, code_entities, code_dependencies
- System: event_queue, file_history, codebase_analysis
- Tech: JSONB metadata, GIN indexes, Row-level security (RLS)
AI Integration
- Claude Code (MCP protocol)
- AI task decomposition
- Pattern recognition
- Risk assessment
- โก Query Speed: <10ms with GIN indexes
- ๐ Real-time: 1s polling interval
- ๐๏ธ Storage: Auto-cleanup processed events (24h)
- ๐ Scalability: Tested with 100+ tasks
- ๐ Analysis: Non-blocking (delegated to Claude)
- ๐ฅ Users: Multi-PM, multi-developer ready
- โ Environment Variables - No hardcoded secrets
- โ Supabase RLS - Row-level security policies
- โ Complete Audit Trail - Every decision recorded
- โ Event Processing - Prevents duplicate actions
- โ Local First - All data in your Supabase instance
- โ GDPR Ready - Export & delete capabilities
- QUICK_START.md - 5-minute setup guide
- PM_GUIDE.md - For Product Managers (Coming Soon)
- DEV_GUIDE.md - For Developers (Coming Soon)
- EXAMPLES.md - Real-world usage examples
- WORKFLOW.md - Complete workflow explanation
- ARCHITECTURE.md - Technical architecture
- API.md - MCP Tools reference
- 60 MCP tools fully functional and tested
- Real-time dashboard with Kanban
- AI story decomposition with dependencies
- Pattern learning & failure detection
- Dependency tracking & conflict detection
- Task metadata (assignee, priority, tags, category)
- Complete audit trail with task history
- Project configuration management
- Claude Code agent synchronization
- AI-powered agent and tool suggestions
- Non-technical PM dashboard view
- Story templates for common features
- Progress reporting & exports
- Stakeholder notifications
- Risk explanations in plain English
- Multi-team workspaces
- Cross-project pattern sharing
- Velocity tracking & estimation
- Auto-assignment based on expertise
- Slack/Teams integration
- LangGraph auto-orchestration
- Predictive risk detection
- Auto-conflict resolution
- Code review automation
- Documentation generation
We welcome contributions from PMs, Developers, and AI enthusiasts!
For Product Managers:
- ๐ Share user story templates
- ๐ก Suggest PM-friendly features
- ๐ Report UX issues
For Developers:
- ๐ง Submit bug fixes
- โจ Add new MCP tools
- ๐ Improve pattern detection
How to contribute:
- Fork the repo
- Create feature branch (
git checkout -b feature/amazing-feature
) - Commit changes (
git commit -m 'Add amazing feature'
) - Push to branch (
git push origin feature/amazing-feature
) - Open Pull Request
- โ 60 MCP Tools - Expanded from 27 to 60 production-ready tools
- โ Automatic Task Analysis - decompose_story now auto-prepares analysis prompts (autoAnalyze=true default)
- โ Project Configuration System - Complete tech stack, agents, tools management
- โ Claude Code Agent Sync - Automatic sync with .claude/agents/ directory
- โ AI Agent/Tool Suggestions - Smart recommendations for tasks
- โ Task History & Events - Complete audit trail with 13 history tools
- โ User Story Health - Monitor completion and status alignment
- โ Bug Fix - Resolved SQL error in get_project_configuration
- โ Full Test Coverage - All 60 tools tested and verified (96.7% success)
- โ Rebranded to Orchestro - "Your AI Development Conductor"
- โ Pattern Analysis Tools - 5 new MCP tools for failure detection
- โ Pattern Frequency - Automatic tracking with database triggers
- โ Risk Assessment - detect_failure_patterns & check_pattern_risk
- โ Task Metadata - assignee, priority, tags fields
- โ PM-focused Documentation - Updated for product owners
- โ New workflow: MCP orchestrates, Claude Code analyzes
- โ 3 execution tools: prepare, save_analysis, get_execution_prompt
- โ tasks.metadata JSONB column
- โ Event queue updated (8 event types)
- โ Guardian verification passed
- Initial MCP implementation
- Basic task management
- AI story decomposition
- Knowledge base integration
"As a PM, I finally understand what developers are building in real-time. Orchestro bridges the gap between product vision and technical implementation." โ Your testimonial here
"Pattern learning saved us hours. The system warned about a risky approach before we wasted time on it." โ Your testimonial here
- ๐ง Issues: GitHub Issues
- ๐ฌ Discussions: GitHub Discussions
- ๐ Docs: Comprehensive guides in the repo
- ๐ Dashboard: http://localhost:3000
- ๐ฆ Twitter: @orchestro_ai (Coming Soon)
MIT License - See LICENSE file for details
- Claude Code - MCP platform & AI development
- Supabase - Database & real-time infrastructure
- Next.js - React framework for dashboard
- React Flow - Beautiful dependency graphs
- shadcn/ui - Gorgeous UI components
Transform product ideas into production code with AI orchestration
Get Started ยท PM Guide ยท Dev Guide ยท See Examples
Status: โ Production Ready (96.7% Test Coverage) Version: 2.2.0 MCP Tools: 60 Made for: PMs ยท Developers ยท Claude Code
๐ผ Conducting development, one task at a time
Made with โค๏ธ by developers who care about product
โญ Star us on GitHub to support the project!