claude-on-django 0.2.0

Autonomous development framework for the Django-Bolt ecosystem

Claude-on-Django

An autonomous development framework that leverages AI agent teams for Django development. Developers describe what they want to build, and the swarm coordinates implementation across all Django layers automatically.

Ported from claude-on-rails with full feature parity.

How It Works

Claude-on-Django creates a team of specialized AI agents, each focused on a specific domain:

Agent	Responsibility
Architect	System design, coordination, project structure
Modeler	Django models, migrations, database optimization
API Specialist	REST API (DRF), JWT/OAuth authentication, serialization
Views	Templates, class-based views, UI (full-stack only)
Services	Business logic, service layer, design patterns
QA Engineer	Unit + integration tests, TDD, test coverage
DevOps	Docker, CI/CD, deployment, production config
Tasks	Celery, background jobs (when Celery detected)

Installation

pip install claude-on-django

Then initialize in your Django project:

python manage.py claude_swarm_init

This will:

• Analyze your project (detect DRF, Celery, GraphQL, HTMX, database, deployment)
• Generate agent prompts in .claude-on-django/prompts/
• Build a swarm topology matching your project features
• Create/update CLAUDE.md with project context

Usage

Natural Language Development

Describe what you want and the swarm handles implementation across all layers:

claude "Add user authentication with JWT and OAuth"
claude "Create a task management API with filtering and pagination"
claude "Optimize the dashboard queries for performance"

Generate a Sandbox Project

Generate a complete example project with CRUD, JWT, OAuth, and full test coverage:

# Generates sandbox using Claude AI agents (requires ANTHROPIC_API_KEY with credits)
python manage.py claude_sandbox_generate --path examples/sandbox

The sandbox includes:

• User registration and profiles
• JWT authentication (SimpleJWT)
• OAuth login (Google, Apple)
• Project and Task CRUD with owner isolation
• Task filtering (status, priority, assignee)
• Task assignment and completion endpoints
• 23 unit tests (models, services, auth)
• 33 integration tests (all API endpoints, pagination, data isolation)

Architecture

claude_on_django/
├── project_analyzer.py      # Detects Django project features
├── swarm_builder.py         # Generates agent topology from analysis
├── configuration.py         # Global settings with env var overrides
├── orchestrator/
│   ├── swarm.py             # SwarmOrchestrator (PydanticAI agents)
│   └── maintainer.py        # CLAUDE.md auto-updater
├── toolsets/
│   └── core.py              # Filesystem and command toolsets
├── mcp_server/
│   ├── structure.py         # MCP server for API metadata scanning
│   └── support.py           # MCP availability and installation
├── agents/                  # 8 agent system prompts (.md)
└── management/commands/
    ├── claude_swarm_init.py          # Project analysis + setup
    └── claude_sandbox_generate.py    # Sandbox project generator

ProjectAnalyzer

Detects your Django project features:

• Frameworks: DRF, GraphQL (Graphene/Strawberry), HTMX, Channels
• Background tasks: Celery
• Database: PostgreSQL, MySQL, SQLite
• Deployment: Docker, Kubernetes, Heroku, Fly.io
• Custom patterns: services, serializers, permissions, middleware, signals

SwarmBuilder

Generates a conditional agent topology based on analysis:

• API-only projects skip the Views agent
• Celery projects get a Tasks agent
• GraphQL projects get a GraphQL agent
• Architect connections are wired to all active agents

Project Structure After Init

your-django-project/
├── CLAUDE.md                              # Project context for Claude
└── .claude-on-django/
    ├── context.md                         # Detected project features
    ├── sessions/                          # Agent session logs
    └── prompts/                           # Agent system prompts
        ├── architect.md
        ├── modeler.md
        ├── api_specialist.md
        ├── views.md
        ├── services.md
        ├── qa_engineer.md
        ├── devops.md
        └── tasks.md

Configuration

Environment Variables

# Required: Anthropic API key (for AI sandbox generation, the default)
ANTHROPIC_API_KEY=sk-ant-...

# Optional: Override the default model (default: anthropic:claude-sonnet-4-6)
CLAUDE_ON_DJANGO_MODEL=anthropic:claude-sonnet-4-6

# Optional: Enable experimental agent teams
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1

Python Configuration

from claude_on_django import configure

configure(
    default_model="anthropic:claude-opus-4-6",
    vibe_mode=False,
    max_parallel_agents=3,
)

Development

# Install dev dependencies
make dev-install

# Run package tests (21 tests)
make test

# Generate sandbox and run its tests (56 tests)
make sandbox-init
make sandbox-test

# Run swarm verification script
make sandbox-run

# Lint
make lint

# Build and publish
make build
make publish

License

MIT License - see LICENSE for details.