A multi-agent AI team that scans any project and generates a complete, structured briefing — ready to be executed by Claude Code or any AI assistant. No API key required.
📦 npm: https://www.npmjs.com/package/jay-crew
╔══════════════════════════════════════════════════╗
║ JAY CREW — Multi-Agent AI Team ║
╚══════════════════════════════════════════════════╝
Jay Crew is a CLI tool that scans a software project and generates a crew-context.md file containing:
- Full project snapshot (tree, config files, dependencies, source samples)
- Automatic technology stack detection (200+ frameworks supported)
- Monorepo and multi-service detection
- Relevant agent definitions for the requested task
- Activation instructions for the Orchestrator
You paste that file into Claude Code (or any AI assistant), say "Run the Jay Crew", and the AI acts as the Orchestrator — running each specialist's X-Ray and synthesizing a complete execution plan.
No API key. No external calls. Jay Crew is the framework; you bring the AI.
# Run directly with npx (no installation needed)
npx jay-crew@latest -p ~/my-project "Full technical analysis"$ npx jay-crew@latest -p ~/my-project "Full analysis"
Need to install the following packages:
jay-crew@0.1.3
Ok to proceed? (y) y
╔══════════════════════════════════════════════════╗
║ JAY CREW — Multi-Agent AI Team ║
╚══════════════════════════════════════════════════╝
📁 Project: /Users/you/my-project
💬 Request: "Full analysis"
🔍 Scanning target project...
✅ 234 files scanned in 0.1s — 133 files in context (35 full · 98 skel) · 100 KB used
🔬 Stack detected: Java · Spring Boot, Docker, PostgreSQL
📦 Monorepo with 3 services detected
🧠 Crew selected: software-architect, backend-dev, devops, qa, engine
📖 Loading agent definitions...
✅ Context file saved: crew-context-2024-01-15T10-30-00.md
─────────────────────────────────────────────────
Paste this file into Claude Code and say:
"Run the Jay Crew on this context."
─────────────────────────────────────────────────
npm install -g jay-crew
jay-crew -p ~/my-app "Add authentication with JWT"Jay Crew automatically detects the technology stack of any project:
🔬 Stack detected: TypeScript (React), Go · Gin, Tailwind CSS, Next.js, PostgreSQL
📦 Monorepo with 3 services detected
🧠 Crew selected: software-architect, backend-dev, frontend-dev, devops, qa
Languages (80+ extensions)
| Category | Languages |
|---|---|
| JVM | Java, Kotlin, Scala, Groovy, Clojure |
| Systems | Go, Rust, C, C++, Zig, Nim, D |
| Scripting | Python, Ruby, PHP, Perl, Lua |
| .NET | C#, F#, Visual Basic |
| Functional | Elixir, Erlang, Haskell, OCaml, Elm, PureScript |
| Data/Scientific | R, Julia, MATLAB |
| Enterprise | COBOL, Fortran, Ada, ABAP |
| Frontend | TypeScript, JavaScript, Vue, Svelte, Astro |
| Mobile | Swift, Kotlin, Dart |
Frameworks & Tools
| Category | Frameworks |
|---|---|
| Java | Spring Boot, Quarkus, Micronaut, Dropwizard, Vert.x, Helidon, Jakarta EE |
| Go | Gin, Echo, Fiber, Chi, Gorilla Mux, Beego |
| Rust | Actix, Axum, Rocket, Warp, Tauri, Yew, Leptos |
| Python | Django, FastAPI, Flask, Tornado, Pyramid, Celery, Airflow |
| Node.js | Express, NestJS, Fastify, Hapi, Koa, AdonisJS, Strapi |
| Ruby | Rails, Sinatra, Hanami, Grape |
| PHP | Laravel, Symfony, CakePHP, WordPress, Drupal |
| .NET | ASP.NET Core, Blazor, MAUI, Entity Framework |
| Elixir | Phoenix, LiveView, Nerves, Absinthe |
| Frontend | React, Vue, Angular, Svelte, Solid, Qwik, Astro, htmx |
| Mobile | Flutter, React Native, Expo, Capacitor, Ionic |
| Styling | Tailwind CSS, Styled Components, Emotion |
| State | Redux, Zustand, MobX, Recoil, Jotai, TanStack Query |
| Testing | Jest, Vitest, Cypress, Playwright |
| ORM/Data | Prisma, TypeORM, Sequelize, SQLAlchemy, Hibernate, Drizzle |
| Queues | Kafka, RabbitMQ, BullMQ, NATS, AWS SQS |
| Databases | PostgreSQL, MySQL, MongoDB, Redis, Elasticsearch, Neo4j, DynamoDB |
| Auth | Auth0, Keycloak, Okta, Clerk, NextAuth, Passport |
Infrastructure & DevOps
| Category | Tools |
|---|---|
| Containers | Docker, Docker Compose, Podman |
| Orchestration | Kubernetes, Helm, Kustomize, Skaffold |
| IaC | Terraform, Pulumi, Ansible, CloudFormation, AWS CDK |
| CI/CD | GitHub Actions, GitLab CI, Jenkins, Azure Pipelines, CircleCI, ArgoCD, FluxCD, Tekton |
| Cloud | Vercel, Netlify, Fly.io, Railway, Heroku, Cloudflare Workers |
| Serverless | AWS Lambda, Azure Functions, Google Cloud Functions, Serverless Framework |
| Observability | Prometheus, Grafana, Datadog, Sentry, OpenTelemetry, Jaeger |
| Monorepo | Nx, Lerna, Turborepo, pnpm Workspaces |
Step 1 ──► Run: npx jay-crew -p ~/my-app "your task"
│
▼
Step 2 ──► Jay Crew scans the project locally
Detects stack, services, and structure
│
▼
Step 3 ──► Auto-selects specialists based on:
• Keywords in your request
• Detected technology stack
• Persona (if specified)
│
▼
Step 4 ──► Generates crew-context-{timestamp}.md
│
▼
Step 5 ──► You paste into Claude Code and say:
"Run the Jay Crew on this context"
│
▼
Step 6 ──► AI acts as the Orchestrator
Runs each specialist's X-Ray
Produces the final Execution Plan
Jay Crew automatically caches project scans to speed up subsequent runs.
On the first run, Jay Crew scans your project and saves the results to .jay-crew-cache/:
$ npx jay-crew -p ~/my-project "Analyze this"
🔍 Scanning target project...
✅ 234 files scanned in 2.1s
💾 Cache saved to .jay-crew-cache/On subsequent runs (if the project hasn't changed), Jay Crew loads from cache instantly:
$ npx jay-crew -p ~/my-project "Different task"
⚡ Cache hit! Loading previous analysis...
🔬 Stack detected: Java · Spring Boot, Docker, PostgreSQL- Location:
.jay-crew-cache/inside your project (not in node_modules) - Contents: Project structure hash, detected stack, file list
- Smart invalidation: Cache is automatically refreshed if any key files change:
package.json,pom.xml,build.gradle,go.mod,Cargo.toml,pyproject.toml- Docker config, CI/CD workflows, dependency files
- Safe: Add
.jay-crew-cache/to.gitignore(already done by default)
To force a full re-scan without using the cache:
npx jay-crew --no-cache -p ~/my-project "Force re-scan"| Agent | Expertise |
|---|---|
| Orchestrator | Coordinates the crew, decides specialists, resolves conflicts, synthesizes the final plan |
| Radar | Real-time research — technologies, versions, trends, best practices |
| Engine | Deep logic — algorithms, code quality, edge cases, design patterns |
| Canvas | Creativity — user flows, UI components, user stories, product strategy |
| Agent | Focus |
|---|---|
product-owner |
Requirements, acceptance criteria, MoSCoW, Definition of Done |
business-analyst |
Business processes, rules, operational flows, entity mapping |
software-architect |
System architecture, C4 diagrams, ADRs, scalability |
backend-dev |
APIs, database schema, auth, server logic — any backend stack |
frontend-dev |
Web UI — any frontend framework |
mobile-dev |
Mobile apps — React Native, Flutter, native iOS/Android |
data-engineer |
Database design, migrations, query optimization, data pipelines |
devops |
Docker, CI/CD, Kubernetes, IaC, observability |
security |
OWASP Top 10, dependency vulnerabilities, auth security, compliance |
qa |
Testing strategy, coverage analysis, quality red flags |
tech-writer |
README, API docs, developer guides, changelogs |
ai-ml |
LLM integration, embeddings, RAG, ML pipelines — when AI adds value |
performance |
Profiling, caching, query optimization, scalability |
Use --persona (or -r) to shape how the Orchestrator presents its analysis.
| Persona | Best for | Auto-boost specialists |
|---|---|---|
new-dev |
Developer new to the codebase | +tech-writer |
senior-dev |
Refactor audit or deep technical review | +performance |
tech-migrator |
Planning a stack or framework migration | +radar, +devops |
task-executor |
Preparing context for an AI agent task | +qa |
tech-lead |
Onboarding material or team briefing | +security, +devops |
due-diligence |
Evaluating a legacy or external codebase | +security, +qa, +devops, +radar |
npx jay-crew [options] "your request"| Flag | Alias | Description |
|---|---|---|
--project <path> |
-p |
Path to the target project (default: current directory) |
--specialists <list> |
-s |
Comma-separated list of specific specialists to force |
--persona <type> |
-r |
Persona profile to shape the Orchestrator output |
--no-cache |
Bypass cache and re-scan the project | |
--help |
-h |
Show help message |
# Analyze any project — stack is auto-detected
npx jay-crew --project ~/my-project "Full technical analysis"
# Force specific specialists
npx jay-crew -p ~/my-app -s backend-dev,security,data-engineer "Add user audit logs"
# Use personas for different analysis styles
npx jay-crew -p ~/my-app --persona new-dev "Explain the codebase"
npx jay-crew -p ~/my-app --persona senior-dev "Optimize performance"
npx jay-crew -p ~/my-app --persona due-diligence "Full technical audit"✅ 234 files scanned in 0.1s — 133 files in context (35 full · 98 skel) · 100 KB used
🔬 Stack detected: Java · Spring Boot, Docker, PostgreSQL
📦 Monorepo with 3 services detected
🧠 Crew selected: software-architect, backend-dev, devops, qa, engine
✅ Context file saved: crew-context-{timestamp}.md
Then paste it into Claude Code and say:
"Run the Jay Crew on this context."
Jay Crew selects specialists using three layers:
| Keywords | Specialists Added |
|---|---|
| auth, login, jwt, oauth | backend-dev, security |
| database, schema, migration | data-engineer |
| docker, kubernetes, ci/cd | devops |
| test, coverage, quality | qa |
| performance, optimize | performance |
| Detected Stack | Specialists Added |
|---|---|
| Java, Go, Python, Node.js | backend-dev |
| React, Vue, Angular | frontend-dev |
| Flutter, React Native | mobile-dev |
| Docker, Kubernetes | devops |
| SQL files, Prisma | data-engineer |
| Persona | Auto-adds |
|---|---|
due-diligence |
security, qa, devops, radar |
tech-migrator |
radar, devops |
tech-lead |
security, devops |
jay-crew/
├── agents/
│ ├── core/
│ │ ├── orchestrator.md ← Coordination + conflict resolution
│ │ ├── radar.md ← Research & validation
│ │ ├── engine.md ← Deep logic & programming
│ │ └── canvas.md ← UX & product strategy
│ └── specialists/ ← 13 specialist agents
├── src/
│ ├── tools/
│ │ ├── project-scanner.ts ← Stack detection (200+ techs)
│ │ └── path-utils.ts
│ ├── types/
│ │ └── index.ts
│ └── index.ts ← CLI entry point
├── bin/
│ └── jay-crew.js ← npx entry point
└── package.json
- No API calls — Jay Crew only reads the filesystem; the AI runs externally
- Technology agnostic — Each agent detects and adapts to the project's stack
- Automatic stack detection — 80+ languages, 200+ frameworks
- Monorepo aware — Identifies and classifies services
- Smart crew selection — Keywords + stack + personas
- Agent definitions as markdown — Easy to read and extend
- Bring your own AI — Works with Claude, ChatGPT, Gemini, or any AI
- Node.js 18+
MIT License — see LICENSE for details.
Contributions are welcome! Feel free to open issues or submit pull requests.
Made with ❤️ by Jeferson Siqueira