AI Agents: Start with AGENT_ORIENTATION.md for roles, authority hierarchy, and entry workflow.
A universal framework for AI-assisted software development workflows.
AIDE provides battle-tested processes, agent primers, and documentation templates for teams building software with AI assistance. Language-agnostic, framework-agnostic, and adaptable to any project type.
AIDE (AI-Assisted Development Environment) is a comprehensive framework that standardizes how AI agents collaborate with human developers on software projects. It provides:
- Process Documentation: Contribution workflows, testing policies, documentation standards
- AI Agent Primers: Role-specific instructions for implementation, review, design, and documentation agents
- Language-Agnostic Templates: Customizable standards for any tech stack
- Battle-Tested Patterns: Proven workflows from real-world AI-assisted development
- Web applications (React, Vue, Angular, etc.)
- Backend services (Node.js, Python, Rust, Go, etc.)
- Mobile apps (React Native, Flutter, Swift, Kotlin, etc.)
- Desktop applications (Electron, Tauri, Qt, etc.)
- Game development (Unity, Godot, Unreal, etc.)
- CLI tools, libraries, frameworks, and more
- Solo developers with AI assistants
- Small teams (2-10 developers)
- Growing teams (10-50 developers)
- Enterprise teams (50+ developers)
AIDE works with any programming language and technology stack. Examples provided for:
- Godot/GDScript (game development)
- Node.js/TypeScript (web/backend)
- Python (ML/backend/scripting)
- Rust (systems/performance)
# Option A: Clone as standalone repo
git clone <your-aide-repo-url> aide-framework
# Option B: Add as git submodule to existing project
cd your-project
git submodule add <your-aide-repo-url> .aideBrowse docs/examples/ for your stack:
godot-gdscript/- Game development with Godotnodejs-typescript/- Web/backend with Node.jspython/- Backend/ML with Pythonrust/- Systems programming with Rust
Copy relevant templates to your project:
# Core process docs (usually symlink these)
ln -s .aide/docs/core/DOCUMENTATION_POLICY.md docs/DOCUMENTATION_POLICY.md
ln -s .aide/docs/core/CONTRIBUTING.md docs/CONTRIBUTING.md
# Tech-specific standards (copy and customize)
cp .aide/docs/examples/nodejs-typescript/TESTING_POLICY.md docs/TESTING_POLICY.md
cp .aide/docs/examples/nodejs-typescript/CODING_GUIDELINES.md docs/CODING_GUIDELINES.md
# Agent primers (usually symlink)
ln -s .aide/docs/agents docs/agentsRecommended: Create a configuration mapping table in docs/AGENTS.md (or equivalent):
## AIDE Configuration (Placeholder Mappings)
| AIDE Placeholder | Project Value |
|-----------------|---------------|
| `{{PROJECT_NAME}}` | YourProjectName |
| `{{TECH_STACK}}` | Your stack (e.g., "Node.js + TypeScript") |
| `{{IMPLEMENTATION_STATUS_QUERY}}` | `gh issue list --label "status:in-progress"` (see [GITHUB_QUERIES.md](docs/agents/GITHUB_QUERIES.md)) |
| `{{RUN_ALL_TESTS_COMMAND}}` | npm test |
...This keeps AIDE generic while your project config defines the mappings. Agents resolve placeholders by reading your config first.
Alternative: Copy templates and replace {{PLACEHOLDERS}} directly (see legacy approach in QUICK_START.md).
See QUICK_START.md for detailed configuration instructions.
AIDE/
├── docs/
│ ├── core/ # Universal process docs (symlink these)
│ │ ├── DOCUMENTATION_POLICY.md
│ │ ├── CONTRIBUTING.md
│ │ └── *.template.md # Project-specific templates
│ │
│ ├── standards/ # Customizable standard templates
│ │ ├── TESTING_POLICY.template.md
│ │ ├── CODING_GUIDELINES.template.md
│ │ └── DEVELOPMENT.template.md
│ │
│ ├── agents/ # AI agent primers (symlink these)
│ │ ├── IMPLEMENTATION_START.md
│ │ ├── PR_REVIEW_START.md
│ │ ├── DOC_REVIEW_START.md
│ │ └── ...
│ │
│ └── examples/ # Language-specific examples
│ ├── godot-gdscript/
│ ├── nodejs-typescript/
│ ├── python/
│ └── rust/
│
└── tools/ # Setup automation scripts
├── init-project.sh
└── init-project.bat
AIDE focuses on how teams work together, not which tools they use. The framework adapts to your stack.
AI agents are treated as team members with specific roles, responsibilities, and boundaries.
Every process is documented, repeatable, and produces consistent results.
Document once in the best place. Reference, don't duplicate.
All changes require tests. No exceptions (unless you document why).
AIDE defines specialized AI agent roles:
- Implementation Agent: Builds features following the implementation workflow (Steps 0-10)
- PR Review Agent: Reviews code for quality, architecture, and standards
- Documentation Agent: Reviews docs for accuracy and consistency
- Codebase Review Agent: Performs holistic codebase audits
- Design Spec Agent: Helps prioritize and spec new features
- Design Workshop Agent: Facilitates high-level design exploration
Each role has a dedicated primer in docs/agents/.
Token Optimization: All agent primers follow token economy best practices (see docs/agents/AGENT_TOKEN_ECONOMY.md) to maximize context budget for code analysis and implementation. Target 150-500 tokens per primer depending on complexity.
The canonical implementation workflow lives in the Implementation Agent primer index:
- docs/agents/IMPLEMENTATION_START.md (index)
- docs/agents/implementation/ (stage-scoped step docs)
To avoid duplication and drift, this README does not restate the step-by-step instructions.
AIDE emerged from real-world experience building software with AI agents. Key insights:
- Structured processes prevent AI drift and scope creep
- Clear boundaries (what agents can/cannot do) improve reliability
- Explicit approval gates maintain human control
- Documentation templates ensure consistency across projects
- Language-agnostic patterns allow framework reuse
A roguelike factory survival game built entirely with AI-assisted development using AIDE workflows. Demonstrates:
- Implementation workflow (Steps 0-10)
- PR review process with inline comments
- Documentation review and accuracy checks
- Design workshop for feature planning
See docs/examples/godot-gdscript/ for implementation details.
We welcome contributions! See CONTRIBUTING.md for:
- How to suggest new agent roles
- Language/framework example additions
- Template improvements
- Process refinements
MIT License - see LICENSE for details.
Created by LightForge Labs as a distillation of AI-assisted development best practices.
- Read QUICK_START.md for setup instructions
- Browse
docs/examples/for your tech stack - Check
docs/agents/for agent-specific guidance - Open an issue for questions or suggestions