AstraForge: The Sovereign AI IDE

The first Integrated Development Environment where AI isn't just a copilot—it's your engineering team.

AstraForge is a standalone, sovereign IDE built on Electron and React, designed to host a specialized panel of autonomous AI agents that debate, architect, and implement code changes through consensus.

Status (Feb 2026): 10 bugs found and fixed in the Feb 2026 audit. Electron terminal wired via IPC/PTY — now fully functional in packaged app. 57 test suites, 188 tests — all passing. OpenRouter/Anthropic/OpenAI providers confirmed working end-to-end. See REPAIR_REPORT.md for full details.

Core Features

⚡ 5-Agent Consensus Architecture

Unlike standard chat-bots, AstraForge orchestrates a team of five specialized agents:

1. Nexus (Orchestrator): Manages the workflow and proposes solutions.
2. Vanguard (Security): Audits code for vulnerabilities and security risks.
3. Prism (Product): Ensures user alignment and product value.
4. Helix (AI Systems): Architects robust AI-native patterns.
5. Cipher (Implementation): Writes the actual production code.

⚡ Consensus-Composer

A dedicated visual interface for reviewing, diffing, and applying complex multi-file changes generated by the agent swarm. Watch as your AI team debates, votes, and generates code in real-time.

⚡ Multi-Provider Support

Connect to your preferred LLM providers:

• OpenAI (GPT-4, GPT-4o, etc.)
• Anthropic (Claude 3.5 Sonnet, Claude 3 Opus)
• Grok (xAI)
• OpenRouter (Access 100+ models)
• Ollama (100% Local, 100% Free)
• LM-Studio (Local models with OpenAI-compatible API)

⚡ Sovereign & Private

Your code, your keys, your local environment. AstraForge is designed to run fully local or self-hosted. No data leaves your machine unless you choose to use cloud APIs.

Installation

Option 1: Docker (Recommended)

Deploy the full environment instantly.

# Clone the repository
git clone https://github.com/up2itnow0822/AstraForge-the-App.git
cd AstraForge-the-App

# Create your .env file
cp example.env .env
# Edit .env and add your API keys

# Build and run
docker compose up --build

Access the IDE at http://localhost:3000

Option 2: Development Mode

Run directly from source for development.

# Install dependencies
npm install

# Start development server (Vite + Electron)
npm run dev

Option 3: Production Build

Build the Electron application for distribution.

# Build everything
npm run build

# Create distributable
npm run dist

Configuration

API Keys

Configure your LLM providers in the Settings modal (gear icon) or via environment variables:

# Global API Keys (in .env file)
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GROK_API_KEY=xai-...
OPENROUTER_API_KEY=sk-or-v1-...

# Local Models
OLLAMA_ENDPOINT=http://127.0.0.1:11434
OLLAMA_MODEL=llama3

Agent-Specific Configuration

Each agent can be configured with a different provider and model via the "Models" tab in Settings.

Usage

1. Enter a Task: Type your development task in the "New Task" box
2. Generate: Click "Generate" to start the AI debate process
3. Watch the Debate: Agents will propose, critique, and vote on solutions
4. Review Changes: Once consensus is reached, review generated code in the Composer
5. Apply: Click "Apply Changes to Disk" to write the files

Architecture

AstraForge/
├── src/
│   ├── core/           # Core business logic
│   │   ├── agents/     # LLM Agent implementations
│   │   ├── debate/     # Debate & consensus system
│   │   └── config/     # Configuration management
│   ├── renderer/       # React UI components
│   │   ├── components/ # UI components
│   │   └── api/        # Bridge to server
│   └── main/           # Server & Electron main
├── specs/              # Technical specifications
└── tests/              # Test suites

Development Philosophy

We believe in High-Agency AI. Tools should handle the implementation details while humans direct the architectural intent. AstraForge is the realization of this 'Sovereign AI' vision.

Testing

# Run all tests
npm test

# Run with coverage
npm run test:coverage

# Run in watch mode
npm run test:watch

Linting

# Check for issues
npm run lint

# Auto-fix issues
npm run lint:fix

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT

Status

Current Version: 1.0.0 (Alpha)

✅ What Works (Feb 2026 — Post-Terminal Fix)

• 5-agent consensus debate flow (Nexus, Vanguard, Prism, Helix, Cipher)
• All LLM providers: OpenRouter, OpenAI, Anthropic, Grok, Ollama, LM-Studio
• Synthesis → User Approval Gate → Code Generation pipeline
• Composer view: file diff, apply-to-disk
• Settings modal: per-agent provider/model/key config
• Connection testing (server/web mode)
• Server mode (Express + Socket.io): full-featured
• Electron mode: full debate flow, approval gate, code generation
• Terminal (Electron) — FIXED: XTerminal now spawns a real PTY via node-pty in the main process and streams I/O to the renderer through Electron IPC (terminal:create, terminal:data, terminal:write, terminal:resize, terminal:exit). Works in both packaged Electron builds and dev mode.

⚠️ Known Limitations

• Agent hot-reload: Changing provider/model in Settings requires app restart
• Memory module: Vector DB integration (LanceDB) is a dependency but not yet connected to the UI
• Connection test (Electron): Not available in packaged app; use web mode for testing

See REPAIR_REPORT.md for the full audit and roadmap.

This is an alpha release. Core functionality is implemented and tested. Feedback and contributions welcome!