claude-on-django 0.1.0

Autonomous development framework for the Django-Bolt ecosystem

Claude-on-Django

An autonomous development framework for the Django-Bolt ecosystem. It leverages Claude Code (CLI) and native Agent Teams to orchestrate a swarm of specialized AI agents to build high-performance APIs.

Overview

Claude-on-Django is a 'performance-first' autonomous swarm framework designed specifically for the Django-Bolt ecosystem. It bridges the gap between high-level Django development and the low-level performance requirements of modern, scalable web services.

How It Works

Claude-on-Django creates a team of specialized AI agents:

• Architect: Coordinates development and makes high-level decisions.
• Modeler: Handles Django models, migrations, and async database design.
• API Specialist: Focuses on high-performance endpoints with msgspec and Actix-Web integration.
• Views: Creates UI templates and manages Django-specific frontend assets.
• Services: Implements business logic and the service layer.
• QA Engineer: Ensures comprehensive test coverage using pytest and TDD.
• DevOps: Handles deployment, infrastructure, and CI/CD pipelines.

Each agent works in their specific domain and can collaborate with other agents to implement complex features.

Installation

You can install the library via pip:

pip install claude-on-django

After installation, initialize the framework in your Django project:

python manage.py claude_swarm_init

This will:

• Analyze your project structure.
• Generate a customized swarm configuration in .claude/.
• Create agent-specific system prompts.
• Set up an automated CLAUDE.md maintainer.

Usage

Natural Language Development

Once initialized, just describe what you want to build in the Claude interface:

# In your Django project directory
claude "Add user authentication with JWT and msgspec validation"

The swarm automatically:

• Analyzes your request.
• Delegates to appropriate specialists (Architect, Modeler, API Specialist, etc.).
• Implements across all layers (models, api, services, tests).
• Follows Django-Bolt performance best practices (async ORM, msgspec).

Sample Commands

> Create a shopping cart with Stripe payment integration
[Complex features are automatically broken down and implemented]

> Optimize the dashboard - it's loading too slowly
[Performance improvements utilizing aget() and msgspec]

> Build a RESTful API for our mobile app with Bolt registration
[API development with high-performance routing]

How It's Different

Traditional Django Development with AI

When using AI assistants for Django development, you typically need to:

• Manually coordinate different aspects of implementation.
• Switch contexts between models, views, and tests.
• Manually ensure consistency and performance standards.

Claude-on-Django Approach

With Claude-on-Django, the swarm automatically:

• Creates models with proper validations and associations.
• Implements high-performance APIs with msgspec.
• Adds comprehensive test coverage using pytest.
• Handles security considerations and async database access.
• Optimizes database queries automatically.

Project Structure

After running the initialization, you'll have:

your-django-project/
├── CLAUDE.md                    # Project-specific context for Claude
└── .claude/
    └── agents/                  # Agent-specific system prompts
        ├── architect.md
        ├── modeler.md
        ├── api_specialist.md
        ├── views.md
        ├── services.md
        ├── qa_engineer.md
        └── devops.md

Customization

Agent Prompts

Customize agent behavior by editing prompts in .claude/agents/:

• Add project-specific conventions.
• Include domain knowledge.
• Define coding standards (e.g., mandatory use of async methods).

Enhanced Documentation with MCP

Claude-on-Django includes a structure for an MCP (Model Context Protocol) server to provide your agents with real-time access to technical documentation and project metadata.

• Up-to-date Documentation: Agents access Django-Bolt resources.
• API Metadata: Real-time scanning of api.py registration metadata.
• Consistent Standards: All agents share the same documentation source.

Testing with Sandbox

A sample project called sandbox is included for testing the framework in examples/sandbox/.

make sandbox-init
make sandbox-run

Development

# Install dev dependencies
make dev-install

# Run tests
make test

# Lint
make lint

License

MIT License - see LICENSE for details.