Autonomous software development workflows with 95% less context usage

Stop loading everything. Start loading exactly what you need, when you need it.

What's New • Features • Quick Start • Web Dashboard • Documentation

We've completed a major overhaul that achieved 80,000+ tokens saved across all resources while dramatically improving discoverability and organization.

Key Achievements:

• 383 Resources Indexed (up from 323, +18.6% growth)
• 207+ Cross-References added for improved navigation
• 77 New Example Files extracted for better reusability
• Token Efficiency: 52-83% savings in real-world usage

• 37 files optimized with code examples moved to dedicated fragments
• ~45,000 tokens saved through example externalization
• 77 new example files created with orchestr8:// URI references
• Improved resource reusability and maintenance

• 6 skill families created (Performance, Security, Testing, Observability, IaC, Error Handling)
• 9 pattern families organized (Event-Driven, Database, Architecture, etc.)
• 42 resources reorganized with hierarchical structure
• 207+ cross-references added for seamless navigation
• ~4,145 tokens net savings

• 2 major agents split into core + advanced modules (Python, Rust)
• 5 workflows enhanced with JIT loading (78% average token reduction)
• 7 resources documented with progressive loading strategies
• Token efficiency: 52-83% savings in typical usage patterns

• Total Resources: 383 fragments
• Domain Experts: 147+ specialized agents
• Reusable Skills: 90+ proven techniques
• Design Patterns: 25+ architectural patterns
• Ready-to-Use Examples: 77+ implementation samples
• useWhen Scenarios: 1,675 automated matching scenarios
• Indexed Keywords: 4,036 unique search terms

• JIT Loading: 77-83% reduction in workflow token usage
• Progressive Loading: 52-82% savings for complex agents
• Example Extraction: ~45,000 tokens saved overall
• Total Impact: 80,000+ tokens saved across all resources

I spent six months watching AI agents waste tokens loading knowledge they'd never use.

The breakthrough? Just-in-time expertise.

orchestr8 is a Claude Code plugin that transforms AI-assisted development through intelligent resource loading via Model Context Protocol (MCP). Instead of cramming 200KB of knowledge into every conversation, it loads a lightweight 2KB catalog—then dynamically fetches exactly what's needed.

The result? 95-98% token reduction. Faster responses. Better conversations. Zero wasted context.

Traditional approach loads everything upfront:

• All TypeScript patterns: 15KB
• All database guides: 10KB
• All security practices: 8KB
• Total waste: ~11,000 tokens before you've even started

orchestr8 loads on-demand:

• Lightweight registry: 250 tokens
• Matched expertise: 800 tokens
• Total used: ~1,050 tokens — only what you need

That's 90% savings on every single workflow.

What makes orchestr8 different:

🔍 Dynamic Resource Matching Semantic fuzzy matching finds relevant expertise automatically based on queries, tags, and capabilities. 1,675 useWhen scenarios and 4,036 indexed keywords ensure precision.

⚡ Multi-Source Providers

• Local: Your custom private resources (fastest, offline-capable)
• AITMPL: 400+ community components (proven patterns)
• GitHub: Your team/company repositories (version-controlled)

🎯 JIT Loading & Progressive Assembly Workflows load ~2KB upfront, fetch 50KB+ on-demand. Registry-first architecture with four optimization modes:

• index: 200-500 tokens (95-98% reduction)
• minimal: 300-500 tokens (ultra-compact JSON)
• catalog: 1,500-2,000 tokens (full metadata)
• full: Complete content when you need it

NEW: Progressive Loading enables 52-83% token savings by loading core modules first, then advanced features only when needed.

🤖 Expert AI Agents (147+ Total) Specialized domain experts for TypeScript, Python, Go, Rust, React, FastAPI, and more—loaded dynamically based on your project. New modular architecture splits complex agents into core + advanced modules.

🧩 Resource Fragments (383 Total) Composable knowledge pieces organized into families:

• Skills: 90+ reusable techniques with hierarchical families
• Patterns: 25+ architectural patterns with 207+ cross-references
• Examples: 77+ ready-to-use implementation samples
• Guides: Step-by-step implementation instructions

💨 Smart Caching LRU cache with configurable TTL: 1hr for prompts, 4-7 days for resources. Sub-millisecond response on cache hits.

🔥 Hot Reload Watch mode with automatic reload during development. Instant feedback loop.

📊 Health Monitoring Real-time provider health, statistics dashboard, and comprehensive observability.

Real-world example:

Task: Build TypeScript REST API with JWT authentication
Without orchestr8:
- Load all TypeScript resources: 15KB
- Load all API patterns: 12KB
- Load all security guides: 8KB
- Load all database patterns: 10KB
Total: 45KB (~11,250 tokens) 😱
With orchestr8 (progressive loading):
- Query registry: 250 tokens
- Load typescript-core agent: 600 tokens
- Load security-auth-jwt skill: 400 tokens
- JIT fetch express-jwt-auth example: 350 tokens
Total: ~1,600 tokens ✅
Savings: 86% reduction! 🚀
With v8.0 optimizations:
- 77 extracted examples reduce duplication
- 207+ cross-references improve navigation
- Progressive loading defers advanced features
- Result: Even more efficient context usage

• Claude Code (latest version)
• Node.js ≥ 18.0.0 (manual installation only)
• npm ≥ 9.0.0 (manual installation only)

The easiest way to get started:

# Step 1: Add the orchestr8 marketplace
/plugin marketplace add seth-schultz/orchestr8
# Step 2: Install the orchestr8 plugin
/plugin install orchestr8@seth-schultz
# Step 3: Verify installation
/help
# You should see /orchestr8:* commands listed

Interactive Installation:

/plugin
# Select "Browse Plugins" → Search for "orchestr8" → Click "Install"

Plugin Management:

# Enable/disable
/plugin enable orchestr8@seth-schultz
/plugin disable orchestr8@seth-schultz
# Uninstall
/plugin uninstall orchestr8@seth-schultz

For development or contributing:

# Clone and build
git clone https://github.com/seth-schultz/orchestr8.git
cd orchestr8/plugins/orchestr8
npm install
npm run build
npm test
# Link to Claude Code settings (.claude/settings.json)
{
  "mcpServers": {
    "orchestr8": {
      "command": "node",
      "args": ["/absolute/path/to/orchestr8/plugins/orchestr8/dist/index.js"]
    }
  }
}

For teams, add to .claude/settings.json for automatic installation:

{
  "plugins": {
    "marketplaces": ["seth-schultz/orchestr8"],
    "installed": ["orchestr8@seth-schultz"]
  }
}

# In Claude Code, try this:
/orchestr8:new-project Build a TypeScript REST API with JWT authentication
# What happens:
# 1. ✅ Matches relevant resources (typescript-developer, security-auth-jwt)
# 2. ✅ Loads only needed fragments (~2KB total)
# 3. ✅ Assembles complete implementation plan
# 4. ✅ Provides step-by-step guidance
# Total tokens used: ~2,000 vs traditional ~11,000 (82% savings!)

See Usage Guide for complete command documentation.

orchestr8 includes a powerful built-in web dashboard that runs automatically with the MCP server. No separate process needed.

# Development mode (HTTP only, hot reload)
npm run dev:http
# Production mode (stdio + HTTP dual transport)
npm run start:http
# Custom port
ORCHESTR8_HTTP_PORT=8080 npm run start:http
# Open browser
https://localhost:1337  # Default port

📊 Overview & Real-Time Stats

Monitor server performance, request metrics, and system health at a glance.

• Live server statistics (uptime, requests, latency)
• Performance charts (P50/P95/P99 latency, cache hit rates)
• Resource utilization metrics
• Real-time WebSocket updates

🧪 Interactive Resource Testing

Test dynamic resource matching with live queries and instant results.

• Execute MCP protocol requests interactively
• Test fuzzy matching with different modes (index, minimal, catalog, full)
• Preview resource content before loading
• Experiment with query parameters and filters
• View protocol-level request/response details

🗂️ Resource Explorer

Browse all available agents, skills, patterns, examples, and guides.

• Category-based navigation
• Real-time search and filtering
• Resource metadata preview
• Token cost estimation
• Quick URI copy for workflows

⏱️ Activity Timeline

Monitor all MCP requests in real-time with detailed inspection.

• Live request stream with timestamps
• Request type and URI details
• Success/failure status
• Response time metrics
• Full request/response payload inspection

🏥 Provider Health Monitoring

Track multi-source provider performance and health.

• Provider status (local, AITMPL, GitHub)
• Response times and success rates
• Cache hit/miss ratios
• Health check results
• Provider-specific statistics

orchestr8 v8.1.0 introduces comprehensive token efficiency monitoring to track, analyze, and optimize your token usage in real-time.

• Real-time Tracking: Monitor token usage on every resource load with <5ms overhead
• Efficiency Metrics: See exactly how much orchestr8's JIT loading saves (typically 95-98%)
• Cost Analysis: Track costs in USD with automatic savings calculation based on Claude Sonnet 4.5 pricing
• Category Breakdown: Understand efficiency by resource type (agents, skills, patterns, etc.)
• Trend Detection: Identify improving, stable, or declining performance patterns over time
• 6 REST API Endpoints: Full programmatic access to all metrics

Access Metrics via API:

# Get current efficiency snapshot
curl https://localhost:1337/api/tokens/efficiency
# View cost savings
curl https://localhost:1337/api/tokens/cost-savings
# Category-based metrics
curl https://localhost:1337/api/tokens/by-category

Example Response:

{
  "overall": {
    "efficiencyPercentage": 96.67,
    "tokensSaved": 435000,
    "costSavingsUSD": 1.35
  },
  "byCategory": [
    {
      "category": "agents",
      "efficiency": 95.2,
      "tokensSaved": 607500,
      "costSavingsUSD": 1.82
    }
  ],
  "trend": {
    "direction": "improving",
    "efficiencyChange": 1.2
  }
}

Available Endpoints:

• GET /api/tokens/efficiency - Comprehensive snapshot with all metrics
• GET /api/tokens/summary - Quick summary for dashboards
• GET /api/tokens/by-category - Category breakdown
• GET /api/tokens/cost-savings - Cost savings report
• GET /api/tokens/trends - Trend analysis
• GET /api/tokens/sessions/:id - Session-level details

Example Workflow:

Task: Build TypeScript REST API with JWT authentication
Without orchestr8:
- Load all TypeScript resources: 15KB (~3,750 tokens)
- Load all API patterns: 12KB (~3,000 tokens)
- Load all security guides: 8KB (~2,000 tokens)
Total: 35KB (~8,750 tokens) - Cost: $0.026
With orchestr8:
- Load typescript-core agent: 2.4KB (~600 tokens)
- Load security-auth-jwt skill: 1.6KB (~400 tokens)
- JIT fetch example: 1.2KB (~300 tokens)
Total: 5.2KB (~1,300 tokens) - Cost: $0.004
Savings: 85% tokens, $0.022 USD ✅

Monthly Impact (100 workflows/month):

• Traditional approach: $2.60/month
• orchestr8 approach: $0.40/month
• Savings: $2.20/month (85%)

Enable/configure token tracking in your environment:

# Environment variables
ORCHESTR8_TOKEN_TRACKING_ENABLED=true
ORCHESTR8_TOKEN_BASELINE_STRATEGY=no_jit
ORCHESTR8_TOKEN_RETENTION_DAYS=7

Or in orchestr8.config.json:

{
  "tokenTracking": {
    "enabled": true,
    "baselineStrategy": "no_jit",
    "retentionDays": 7,
    "enableTrends": true
  }
}

Complete documentation available:

• Token Efficiency Monitoring - Complete feature guide
• API Reference - REST API documentation

orchestr8 features a powerful provider system for loading resources from multiple sources with intelligent caching and automatic fallback.

Create orchestr8.config.json:

{
  "providers": {
    "local": { "enabled": true },
    "aitmpl": { "enabled": true },
    "github": {
      "enabled": true,
      "repos": ["davila7/claude-code-templates", "mycompany/resources"],
      "auth": { "token": "${GITHUB_TOKEN}", "type": "personal" }
    }
  }
}

Set GitHub Token (optional, 5000 req/hr vs 60):

export GITHUB_TOKEN="ghp_your_personal_access_token"

Workflows automatically search all enabled providers in parallel:

/orchestr8:new-project Build a Rust web server with async
# What happens:
# 1. LocalProvider searches your custom resources
# 2. AITMPLProvider searches 400+ community components
# 3. GitHubProvider searches configured repos
# 4. Results merged by relevance score
# 5. Top resources assembled and cached for 7 days

Example assembled content:

• rust-pro agent from AITMPL (community expert)
• async-patterns skill from Local (your customizations)
• web-server-template from GitHub (company standards)

Token Efficiency:

Without providers: Load all 450 resources = 45MB (~11M tokens)
With providers: Load top 3-5 resources = 5KB (~1,250 tokens)
Savings: 99.99% ✅

Resource Discovery:

• Community-proven patterns from AITMPL
• Company-specific standards from GitHub
• Local customizations and private resources
• Automatic relevance ranking

Reliability:

• Automatic fallback (Local → AITMPL → GitHub)
• Health monitoring with auto-disable
• Offline capability with local provider
• Sub-millisecond cache hits

# Start HTTP server
npm run dev:http
# Open https://localhost:1337
# View: Provider health, statistics, cache rates, real-time updates

API Examples:

# Check all providers
curl https://localhost:1337/api/providers/health/all
# Search across providers
curl "https://localhost:1337/api/search/multi?q=typescript&categories=agent"
# Get statistics
curl https://localhost:1337/api/providers/aitmpl/stats

Complete provider system documentation:

• Provider Overview - Introduction and quick start
• Architecture - Technical design
• Configuration - All options
• Usage Guide - Practical examples
• API Reference - HTTP API docs
• Development Guide - Build custom providers

orchestr8 implements a sophisticated MCP-based architecture optimized for token efficiency through progressive loading and intelligent resource organization.

Prompt Loader Loads workflow prompts with argument substitution. Cached with 1hr TTL.

Resource Loader Resolves static URIs and performs dynamic fuzzy matching. The brain of JIT loading. Now supports:

• Progressive module loading (core → advanced)
• Cross-reference resolution across 207+ links
• Example extraction and URI-based references
• Hierarchical skill family navigation

URI Parser Supports category/resource and match?query=... formats for flexible resource access. Enhanced with:

• orchestr8:// URI scheme for examples
• Cross-reference resolution
• Family-based resource lookup

Fuzzy Matcher Semantic scoring via tags, capabilities, and use-cases with 1,675 useWhen scenarios and 4,036 indexed keywords. Finds the right expertise automatically.

Fragment Assembly Combines fragments within token budget limits with surgical precision:

• 383 Total Fragments organized hierarchically
• 6 Skill Families for structured discovery
• 9 Pattern Families with cross-references
• 77 Example Files for reusable code samples

Smart Caching LRU cache with separate TTLs: prompts (1hr), resources (4-7 days). Sub-millisecond on hits.

Resource Hierarchy (NEW in v8.0)

resources/
├── agents/         147+ domain experts (some with core + advanced modules)
├── skills/         90+ techniques organized in 6 families
├── patterns/       25+ designs organized in 9 families
├── examples/       77+ ready-to-use code samples
├── workflows/      25+ multi-phase processes
└── guides/         Step-by-step implementation docs

See Architecture Documentation for detailed design.

Monthly cost comparison: Traditional approach vs orchestr8

Response time and resource usage across different workflows

orchestr8's fuzzy matching delivers 95%+ relevance vs 60-70% with traditional static loading

cd plugins/orchestr8
# Install dependencies
npm install
# Build TypeScript
npm run build
# Watch mode (auto-rebuild)
npm run watch
# Development mode with hot reload
npm run dev

# Run all tests
npm test
# Run specific test suites
npm run test:unit           # Unit tests only
npm run test:integration    # Integration tests only
npm run test:watch          # Watch mode
# Coverage report
npm run test:coverage
# Verify project structure
npm run verify

# Lint code
npm run lint
# Format code
npm run format
# Type check
npm run type-check
# Clean build artifacts
npm run clean

See Development Guide for complete setup.

All comprehensive documentation is in plugins/orchestr8/docs/

We welcome contributions! orchestr8 is built by developers, for developers.

1. Read the guides:

   • Contributing Guide - Workflow and guidelines
   • Code of Conduct - Community standards
   • Authoring Guide - Create resources

2. Pick an area:

   • 🐛 Fix bugs
   • ✨ Add features
   • 📝 Improve documentation
   • 🧪 Add tests
   • 🎨 Create agents, skills, or patterns

3. Submit a PR:

   • Fork the repository
   • Create a feature branch
   • Make your changes
   • Add tests
   • Submit a pull request

• ✅ Token efficiency first - Every change should maintain or improve token usage
• ✅ Test everything - Comprehensive test coverage required
• ✅ Document as you go - Code without docs is incomplete
• ✅ Security by default - Security is not optional
• ✅ Performance matters - Sub-millisecond response times are the goal

Security is our top priority. orchestr8 implements defense-in-depth security:

• stdio Transport: No network ports, local-only operation
• Process Isolation: MCP server runs in isolated Node.js process
• Input Validation: All queries and URIs sanitized
• Dependency Scanning: Automated npm audit on every PR
• Secret Scanning: Gitleaks prevents credential leaks
• Supply Chain Security: All GitHub Actions pinned to commit SHAs

🚨 DO NOT open public issues for security vulnerabilities.

Instead:

• 📧 Email: security@orchestr8.builders
• 🔒 Use GitHub Security Advisories

Response Timeline:

• ⚡ Initial response: 24 hours
• 🔍 Confirmation: 3 business days
• 🛠️ Fix timeline: 7-60 days (based on severity)

See SECURITY.md for complete policy.

orchestr8 is open source software licensed under the MIT License.

MIT License - Copyright (c) 2024 Seth Schultz
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

See LICENSE for full text.

orchestr8 stands on the shoulders of giants:

• Model Context Protocol - The foundation for dynamic resource loading
• Claude Code - Powering AI-assisted development
• @modelcontextprotocol/sdk - MCP TypeScript implementation
• Open Source Community - For inspiration, feedback, and contributions

⬆ Back to Top

Made with ❤️ by the orchestr8 community

Questions? Check the documentation or open an issue